Search

Spurred by widespread adoption of artificial intelligence and machine learning, “data” is becoming a key production factor, comparable in importance to capital, land, or labour in an increasingly digital economy. In spite of an ever-growing demand for third-party data in the B2B market, firms are generally reluctant to share their information. This is due to the unique characteristics of “data” as an economic good (a freely replicable, non-depletable asset holding a highly combinatorial and context-specific value), which moves digital companies to hoard and protect their “valuable” data assets, and to integrate across the whole value chain seeking to monopolise the provision of innovative services built upon them. As a result, most of those valuable assets still remain unexploited in corporate silos nowadays. This situation is shaping the so-called data economy around a number of champions, and it is hampering the benefits of a global data exchange on a large scale. Some analysts have estimated the potential value of the data economy in US$2.5 trillion globally by 2025. Not surprisingly, unlocking the value of data has become a central policy of the European Union, which also estimated the size of the data economy in 827C billion for the EU27 in the same period. Within the scope of the European Data Strategy, the European Commission is also steering relevant initiatives aimed to identify relevant cross-industry use cases involving different verticals, and to enable sovereign data exchanges to realise them. Among individuals, the massive collection and exploitation of personal data by digital firms in exchange of services, often with little or no consent, has raised a general concern about privacy and data protection. Apart from spurring recent legislative developments in this direction, this concern has raised some voices warning against the unsustainability of the existing digital economics (few digital champions, potential negative impact on employment, growing inequality), some of which propose that people are paid for their data in a sort of worldwide data labour market as a potential solution to this dilemma [114, 115, 155]. From a technical perspective, we are far from having the required technology and algorithms that will enable such a human-centric data economy. Even its scope is still blurry, and the question about the value of data, at least, controversial. Research works from different disciplines have studied the data value chain, different approaches to the value of data, how to price data assets, and novel data marketplace designs. At the same time, complex legal and ethical issues with respect to the data economy have risen around privacy, data protection, and ethical AI practices. In this dissertation, we start by exploring the data value chain and how entities trade data assets over the Internet. We carry out what is, to the best of our understanding, the most thorough survey of commercial data marketplaces. In this work, we have catalogued and characterised ten different business models, including those of personal information management systems, companies born in the wake of recent data protection regulations and aiming at empowering end users to take control of their data. We have also identified the challenges faced by different types of entities, and what kind of solutions and technology they are using to provide their services. Then we present a first of its kind measurement study that sheds light on the prices of data in the market using a novel methodology. We study how ten commercial data marketplaces categorise and classify data assets, and which categories of data command higher prices. We also develop classifiers for comparing data products across different marketplaces, and we study the characteristics of the most valuable data assets and the features that specific vendors use to set the price of their data products. Based on this information and adding data products offered by other 33 data providers, we develop a regression analysis for revealing features that correlate with prices of data products. As a result, we also implement the basic building blocks of a novel data pricing tool capable of providing a hint of the market price of a new data product using as inputs just its metadata. This tool would provide more transparency on the prices of data products in the market, which will help in pricing data assets and in avoiding the inherent price fluctuation of nascent markets. Next we turn to topics related to data marketplace design. Particularly, we study how buyers can select and purchase suitable data for their tasks without requiring a priori access to such data in order to make a purchase decision, and how marketplaces can distribute payoffs for a data transaction combining data of different sources among the corresponding providers, be they individuals or firms. The difficulty of both problems is further exacerbated in a human-centric data economy where buyers have to choose among data of thousands of individuals, and where marketplaces have to distribute payoffs to thousands of people contributing personal data to a specific transaction. Regarding the selection process, we compare different purchase strategies depending on the level of information available to data buyers at the time of making decisions. A first methodological contribution of our work is proposing a data evaluation stage prior to datasets being selected and purchased by buyers in a marketplace. We show that buyers can significantly improve the performance of the purchasing process just by being provided with a measurement of the performance of their models when trained by the marketplace with individual eligible datasets. We design purchase strategies that exploit such functionality and we call the resulting algorithm Try Before You Buy, and our work demonstrates over synthetic and real datasets that it can lead to near-optimal data purchasing with only O(N) instead of the exponential execution time - O(2N) - needed to calculate the optimal purchase. With regards to the payoff distribution problem, we focus on computing the relative value of spatio-temporal datasets combined in marketplaces for predicting transportation demand and travel time in metropolitan areas. Using large datasets of taxi rides from Chicago, Porto and New York we show that the value of data is different for each individual, and cannot be approximated by its volume. Our results reveal that even more complex approaches based on the “leave-one-out” value, are inaccurate. Instead, more complex and acknowledged notions of value from economics and game theory, such as the Shapley value, need to be employed if one wishes to capture the complex effects of mixing different datasets on the accuracy of forecasting algorithms. However, the Shapley value entails serious computational challenges. Its exact calculation requires repetitively training and evaluating every combination of data sources and hence O(N!) or O(2N) computational time, which is unfeasible for complex models or thousands of individuals. Moreover, our work paves the way to new methods of measuring the value of spatio-temporal data. We identify heuristics such as entropy or similarity to the average that show a significant correlation with the Shapley value and therefore can be used to overcome the significant computational challenges posed by Shapley approximation algorithms in this specific context. We conclude with a number of open issues and propose further research directions that leverage the contributions and findings of this dissertation. These include monitoring data transactions to better measure data markets, and complementing market data with actual transaction prices to build a more accurate data pricing tool. A human-centric data economy would also require that the contributions of thousands of individuals to machine learning tasks are calculated daily. For that to be feasible, we need to further optimise the efficiency of data purchasing and payoff calculation processes in data marketplaces. In that direction, we also point to some alternatives to repetitively training and evaluating a model to select data based on Try Before You Buy and approximate the Shapley value. Finally, we discuss the challenges and potential technologies that help with building a federation of standardised data marketplaces. The data economy will develop fast in the upcoming years, and researchers from different disciplines will work together to unlock the value of data and make the most out of it. Maybe the proposal of getting paid for our data and our contribution to the data economy finally flies, or maybe it is other proposals such as the robot tax that are finally used to balance the power between individuals and tech firms in the digital economy. Still, we hope our work sheds light on the value of data, and contributes to making the price of data more transparent and, eventually, to moving towards a human-centric data economy. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Georgios Smaragdakis.- Secretario: Ángel Cuevas Rumín.- Vocal: Pablo Rodríguez Rodríguez

Smartphones have become inherent to the every day life of billions of people worldwide, and they are used to perform activities such as gaming, interacting with our peers or working. While extremely useful, smartphone apps also have drawbacks, as they can affect the security and privacy of users. Android devices hold a lot of personal data from users, including their social circles (e.g., contacts), usage patterns (e.g., app usage and visited websites) and their physical location. Like in most software products, Android apps often include third-party code (Software Development Kits or SDKs) to include functionality in the app without the need to develop it in-house. Android apps and third-party components embedded in them are often interested in accessing such data, as the online ecosystem is dominated by data-driven business models and revenue streams like advertising. The research community has developed many methods and techniques for analyzing the privacy and security risks of mobile apps, mostly relying on two techniques: static code analysis and dynamic runtime analysis. Static analysis analyzes the code and other resources of an app to detect potential app behaviors. While this makes static analysis easier to scale, it has other drawbacks such as missing app behaviors when developers obfuscate the app’s code to avoid scrutiny. Furthermore, since static analysis only shows potential app behavior, this needs to be confirmed as it can also report false positives due to dead or legacy code. Dynamic analysis analyzes the apps at runtime to provide actual evidence of their behavior. However, these techniques are harder to scale as they need to be run on an instrumented device to collect runtime data. Similarly, there is a need to stimulate the app, simulating real inputs to examine as many code-paths as possible. While there are some automatic techniques to generate synthetic inputs, they have been shown to be insufficient. In this thesis, we explore the benefits of combining static and dynamic analysis techniques to complement each other and reduce their limitations. While most previous work has often relied on using these techniques in isolation, we combine their strengths in different and novel ways that allow us to further study different privacy issues on the Android ecosystem. Namely, we demonstrate the potential of combining these complementary methods to study three inter-related issues: • A regulatory analysis of parental control apps. We use a novel methodology that relies on easy-to-scale static analysis techniques to pin-point potential privacy issues and violations of current legislation by Android apps and their embedded SDKs. We rely on the results from our static analysis to inform the way in which we manually exercise the apps, maximizing our ability to obtain real evidence of these misbehaviors. We study 46 publicly available apps and find instances of data collection and sharing without consent and insecure network transmissions containing personal data. We also see that these apps fail to properly disclose these practices in their privacy policy. • A security analysis of the unauthorized access to permission-protected data without user consent. We use a novel technique that combines the strengths of static and dynamic analysis, by first comparing the data sent by applications at runtime with the permissions granted to each app in order to find instances of potential unauthorized access to permission protected data. Once we have discovered the apps that are accessing personal data without permission, we statically analyze their code in order to discover covert- and side-channels used by apps and SDKs to circumvent the permission system. This methodology allows us to discover apps using the MAC address as a surrogate for location data, two SDKs using the external storage as a covert-channel to share unique identifiers and an app using picture metadata to gain unauthorized access to location data. • A novel SDK detection methodology that relies on obtaining signals observed both in the app’s code and static resources and during its runtime behavior. Then, we rely on a tree structure together with a confidence based system to accurately detect SDK presence without the need of any a priory knowledge and with the ability to discern whether a given SDK is part of legacy or dead code. We prove that this novel methodology can discover third-party SDKs with more accuracy than state-of-the-art tools both on a set of purpose-built ground-truth apps and on a dataset of 5k publicly available apps. With these three case studies, we are able to highlight the benefits of combining static and dynamic analysis techniques for the study of the privacy and security guarantees and risks of Android apps and third-party SDKs. The use of these techniques in isolation would not have allowed us to deeply investigate these privacy issues, as we would lack the ability to provide real evidence of potential breaches of legislation, to pin-point the specific way in which apps are leveraging cover and side channels to break Android’s permission system or we would be unable to adapt to an ever-changing ecosystem of Android third-party companies. The works presented in this thesis were partially funded within the framework of the following projects and grants: • European Union’s Horizon 2020 Innovation Action program (Grant Agreement No. 786741, SMOOTH Project and Grant Agreement No. 101021377, TRUST AWARE Project). • Spanish Government ODIO NºPID2019-111429RB-C21/PID2019-111429RBC22. • The Spanish Data Protection Agency (AEPD) • AppCensus Inc. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Srdjan Matic.- Secretario: Guillermo Suárez-Tangil.- Vocal: Ben Stock

Este trabajo se enmarca en el amplio contexto del Internet de las Cosas, y se centra en la interoperabilidad entre humanos y los sistemas de automatización. En cómo los dispositivos que rodean al usuario interactúan con los elementos computacionales que encuentran en su entorno, pudiendo generar, de forma automática, tareas que sean delegables a otros elementos. Siendo más concretos, la intención de este trabajo es doble. En primer lugar, proponer una arquitectura, basada en el uso de microservicios, gracias a la cual poder migrar las tareas presentes en los dispositivos cercanos al usuario a otros dentro del mismo entorno. En segundo lugar, desarrollar un sistema que haga posible la interacción del usuario con la arquitectura, permitiendo dar instrucciones al entorno a través del lenguaje natural. El desarrollo de arquitecturas basadas en microservicios posibilita la migración de un dispositivo a otro de determinadas piezas de software ligeras. Para realizar estas migraciones de la manera más óptima posible, es necesario conocer la compatibilidad entre microservicios y dispositivos. Este trabajo proporciona una manera de determinar esta compatibilidad. La interacción entre un usuario y su entorno a través del lenguaje natural está muy presente en nuestro día a día, por ejemplo, a través de asistentes virtuales de voz. Se trata de un mecanismo que busca facilitar la interacción entre humano y entorno, pero que, en determinadas situaciones, puede no responder de forma acertada a las peticiones del usuario. Una de las razones por las que se pueden dar estas situaciones es la falta de compresión de la petición dada porque no se han utilizado los términos adecuados. Otra de las razones es la falta de una definición más precisa del entorno donde es posible que coexistan diferentes elementos que, aún siendo diferentes, pueden satisfacer de una forma similar los deseos del usuario dando lugar a posibles confusiones. Que un usuario sea capaz de dar órdenes a un entorno sin necesidad de conocer el lenguaje o los patrones necesarios para que un sistema las entienda, aumenta el número de usuarios que son capaces de interactuar con su entorno. En este trabajo se presenta un sistema que pretende eliminar la necesidad de entrenamiento previo del mismo, así como el requisito, por parte del usuario, de conocer los enunciados o palabras clave necesarias para generar órdenes. A continuación se presentan las principales contribuciones en este contexto: Proposición de un conjunto de directrices que forman el marco teórico para el despliegue de una arquitectura basada en microservicios que ayudan a la descentralización del análisis de los diferentes dispositivos que coexisten en ella, así como al análisis de los diferentes microservicios presentes en la misma. Proposición de un conjunto de directrices que forman el marco teórico para el despliegue de infraestructuras efímeras, en arquitecturas basadas en microservicios, que ayudan a simplificar la toma de decisiones tanto en su despliegue como en sus migraciones. En concreto, permitiendo generar entornos donde llevar a cabo diferentes análisis de rendimiento en los dispositivos capaces de alojar un microservicio, añadiendo así un método con el que obtener datos para la toma de decisiones. Proposición de un conjunto de directrices que forman el marco teórico para el análisis de órdenes dadas en lenguaje natural por parte de un usuario humano a su entorno cercano. Proposición de un conjunto de directrices que forman el marco teórico que facilita el desarrollo de servicios compatibles con la utilización del lenguaje natural para su uso. Proposición de un conjunto de directrices que forman el marco teórico que permite, a los desarrolladores, incorporar la información necesaria sobre sus servicios y las acciones que se pueden realizar gracias a ellos, permitiendo una interacción fluida sin añadir complejidad al lenguaje que se debe utilizar. Proposición de un conjunto de directrices que forman el marco teórico que facilita la virtualización de las acciones realizables en un entorno a través del análisis de las descripciones provistas por sus desarrolladores. Proposición de un conjunto de directrices que forman el marco teórico que pretende eliminar la necesidad de entrenamiento, tanto del entorno, como de los usuarios, para poder relacionar la intención del usuario y la acción que pretende llevar a cabo. Implementación, como prueba de concepto, de una arquitectura de microservicios que sigue las directrices expuestas en este trabajo, donde la toma de decisiones sobre sus migraciones responde, tanto a la posibilidad de despliegue de programas ligeros y de corta duración para el análisis de rendimiento, así como a los deseos expuestos por usuarios potenciales a través del lenguaje natural. Parte del contenido se desarrolló como línea de investigación en varios proyectos de I+D, denominados: INRISCO: Monitorización de Incidentes en Comunidades Inteligentes: Seguridad y Movilidad (TEC2014-54335-C4-2-R) MAGOS: Secure Smart Grid Using Open Source Intelligence (TEC2017- 84197-C4-3-R) CYNAMON co-financed by European Structural Funds (ESF end FEDER) (P2018/TCS-4566) Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Carlos García Rubio.- Secretario: José María Flores Arias.- Vocal: Ramona Ruiz Blázquez

Current robotic systems handle a different range of applications such as video surveillance, delivery of goods, cleaning, material handling, assembly, painting, or pick and place services. These systems have been embraced not only by the general population but also by the vertical industries to help them in performing daily activities. Traditionally, the robotic systems have been deployed in standalone robots that were exclusively dedicated to performing a specific task such as cleaning the floor in indoor environments. In recent years, cloud providers started to offer their infrastructures to robotic systems for offloading some of the robot’s functions. This ultimate form of the distributed robotic system was first introduced 10 years ago as cloud robotics and nowadays a lot of robotic solutions are appearing in this form. As a result, standalone robots became software-enhanced objects with increased reconfigurability as well as decreased complexity and cost. Moreover, by offloading the heavy processing from the robot to the cloud, it is easier to share services and information from various robots or agents to achieve better cooperation and coordination. Cloud robotics is suitable for human-scale responsive and delay-tolerant robotic functionalities (e.g., monitoring, predictive maintenance). However, there is a whole set of real-time robotic applications (e.g., remote control, motion planning, autonomous navigation) that can not be executed with cloud robotics solutions, mainly because cloud facilities traditionally reside far away from the robots. While the cloud providers can ensure certain performance in their infrastructure, very little can be ensured in the network between the robots and the cloud, especially in the last hop where wireless radio access networks are involved. Over the last years advances in edge computing, fog computing, 5G NR, network slicing, Network Function Virtualization (NFV), and network orchestration are stimulating the interest of the industrial sector to satisfy the stringent and real-time requirements of their applications. Robotic systems are a key piece in the industrial digital transformation and their benefits are very well studied in the literature. However, designing and implementing a robotic system that integrates all the emerging technologies and meets the connectivity requirements (e.g., latency, reliability) is an ambitious task. This thesis studies the integration of modern Information andCommunication Technologies (ICTs) in robotic systems and proposes some robotic enhancements that tackle the real-time constraints of robotic services. To evaluate the performance of the proposed enhancements, this thesis departs from the design and prototype implementation of an edge native robotic system that embodies the concepts of edge computing, fog computing, orchestration, and virtualization. The proposed edge robotics system serves to represent two exemplary robotic applications. In particular, autonomous navigation of mobile robots and remote-control of robot manipulator where the end-to-end robotic system is distributed between the robots and the edge server. The open-source prototype implementation of the designed edge native robotic system resulted in the creation of two real-world testbeds that are used in this thesis as a baseline scenario for the evaluation of new innovative solutions in robotic systems. After detailing the design and prototype implementation of the end-to-end edge native robotic system, this thesis proposes several enhancements that can be offered to robotic systems by adapting the concept of edge computing via the Multi-Access Edge Computing (MEC) framework. First, it proposes exemplary network context-aware enhancements in which the real-time information about robot connectivity and location can be used to dynamically adapt the end-to-end system behavior to the actual status of the communication (e.g., radio channel). Three different exemplary context-aware enhancements are proposed that aim to optimize the end-to-end edge native robotic system. Later, the thesis studies the capability of the edge native robotic system to offer potential savings by means of computation offloading for robot manipulators in different deployment configurations. Further, the impact of different wireless channels (e.g., 5G, 4G andWi-Fi) to support the data exchange between a robot manipulator and its remote controller are assessed. In the following part of the thesis, the focus is set on how orchestration solutions can support mobile robot systems to make high quality decisions. The application of OKpi as an orchestration algorithm and DLT-based federation are studied to meet the KPIs that autonomously controlledmobile robots have in order to provide uninterrupted connectivity over the radio access network. The elaborated solutions present high compatibility with the designed edge robotics system where the robot driving range is extended without any interruption of the end-to-end edge robotics service. While the DLT-based federation extends the robot driving range by deploying access point extension on top of external domain infrastructure, OKpi selects the most suitable access point and computing resource in the cloud-to-thing continuum in order to fulfill the latency requirements of autonomously controlled mobile robots. To conclude the thesis the focus is set on how robotic systems can improve their performance by leveraging Artificial Intelligence (AI) and Machine Learning (ML) algorithms to generate smart decisions. To do so, the edge native robotic system is presented as a true embodiment of a Cyber-Physical System (CPS) in Industry 4.0, showing the mission of AI in such concept. It presents the key enabling technologies of the edge robotic system such as edge, fog, and 5G, where the physical processes are integrated with computing and network domains. The role of AI in each technology domain is identified by analyzing a set of AI agents at the application and infrastructure level. In the last part of the thesis, the movement prediction is selected to study the feasibility of applying a forecast-based recovery mechanism for real-time remote control of robotic manipulators (FoReCo) that uses ML to infer lost commands caused by interference in the wireless channel. The obtained results are showcasing the its potential in simulation and real-world experimentation. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Karl Holger.- Secretario: Joerg Widmer.- Vocal: Claudio Cicconetti

The Android Open Source Project (AOSP) was first released by Google in 2008 and has since become the most used operating system [Andaf]. Thanks to the openness of its source code, any smartphone vendor or original equipment manufacturer (OEM) can modify and adapt Android to their specific needs, or add proprietary features before installing it on their devices in order to add custom features to differentiate themselves from competitors. This has created a complex and diverse supply chain, completely opaque to end-users, formed by manufacturers, resellers, chipset manufacturers, network operators, and prominent actors of the online industry that partnered with OEMs. Each of these stakeholders can pre-install extra apps, or implement proprietary features at the framework level. However, such customizations can create privacy and security threats to end-users. Preinstalled apps are privileged by the operating system, and can therefore access system APIs or personal data more easily than apps installed by the user. Unfortunately, despite these potential threats, there is currently no end-to-end control over what apps come pre-installed on a device and why, and no traceability of the different software and hardware components used in a given Android device. In fact, the landscape of pre-installed software in Android and its security and privacy implications has largely remained unexplored by researchers. In this thesis, I investigate the customization of Android devices and their impact on the privacy and security of end-users. Specifically, I perform the first large-scale and systematic analysis of pre-installed Android apps and the supply chain. To do so, I first develop an app, Firmware Scanner [Sca], to crowdsource close to 34,000 Android firmware versions from 1,000 different OEMs from all over the world. This dataset allows us to map the stakeholders involved in the supply chain and their relationships, from device manufacturers and mobile network operators to third-party organizations like advertising and tracking services, and social network platforms. I could identify multiple cases of privacy-invasive and potentially harmful behaviors. My results show a disturbing lack of transparency and control over the Android supply chain, thus showing that it can be damageable privacy- and security-wise to end-users. Next, I study the evolution of the Android permission system, an essential security feature of the Android framework. Coupled with other protection mechanisms such as process sandboxing, the permission system empowers users to control what sensitive resources (e.g., user contacts, the camera, location sensors) are accessible to which apps. The research community has extensively studied the permission system, but most previous studies focus on its limitations or specific attacks. In this thesis, I present an up-to-date view and longitudinal analysis of the evolution of the permissions system. I study how some lesser-known features of the permission system, specifically permission flags, can impact the permission granting process, making it either more restrictive or less. I then highlight how pre-installed apps developers use said flags in the wild and focus on the privacy and security implications. Specifically, I show the presence of third-party apps, installed as privileged system apps, potentially using said features to share resources with other third-party apps. Another salient feature of the permission system is its extensibility: apps can define their own custom permissions to expose features and data to other apps. However, little is known about how widespread the usage of custom permissions is, and what impact these permissions may have on users’ privacy and security. In the last part of this thesis, I investigate the exposure and request of custom permissions in the Android ecosystem and their potential for opening privacy and security risks. I gather a 2.2-million-app-large dataset of both pre-installed and publicly available apps using both Firmware Scanner and purpose-built app store crawlers. I find the usage of custom permissions to be pervasive, regardless of the origin of the apps, and seemingly growing over time. Despite this prevalence, I find that custom permissions are virtually invisible to end-users, and their purpose is mostly undocumented. While Google recommends that developers use their reverse domain name as the prefix of their custom permissions [Gpla], I find widespread violations of this recommendation, making sound attribution at scale virtually impossible. Through static analysis methods, I demonstrate that custom permissions can facilitate access to permission-protected system resources to apps that lack those permissions, without user awareness. Due to the lack of tools for studying such risks, I design and implement two tools, PermissionTracer [Pere] and PermissionTainter [Perd] to study custom permissions. I highlight multiple cases of concerning use of custom permissions by Android apps in the wild. In this thesis, I systematically studied, at scale, the vast and overlooked ecosystem of preinstalled Android apps. My results show a complete lack of control of the supply chain which is worrying, given the huge potential impact of pre-installed apps on the privacy and security of end-users. I conclude with a number of open research questions and future avenues for further research in the ecosystem of the supply chain of Android devices. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Douglas Leith.- Secretario: Rubén Cuevas Rumín.- Vocal: Hamed Haddadi

Wireless communication is a transformative technology that has changed the way we work, communicate and enjoy our free time. The number of connected devices is expected to increase to 29.4 billion by 2030. As a result, there is a demand for higher data rates, lower delays and constant connectivity of wireless devices. These demands have driven the networking community to seek new technologies as these requirements are beyond the capabilities of current networks. Communication at higher frequencies, beyond 10 GHz where most of current communications systems operate, could be the game changing technology for the next generation of networks. At millimeter-wave frequencies, 30-300 GHz, the available spectrum is larger than all spectrum currently allocated to cellular and wireless area networks (WLAN). The unlicensed spectrum at 60 GHz alone can offer 10 to 100 times more spectrum than it is available in current unlicensed WLANs. The larger bandwidth allocations allow for increased datarates. These multi-gigabit per second rates and milisecond latencies are now achievable thanks to new directional high gain antennas and cost-effective CMOS technology that can operate at mm-wave frequencies. However, with the use of this new technology new challenges arise. For example, since the communication is directional in mm-wave, the best direction to communicate needs to be decided, and updated in mobility scenarios. Aside from the technical deployment challenges at this frequency, another challenge is how to provide increased capabilities for the network such as localization and passive localization. The specific characteristic of propagation at mm-wave frequencies such as the quasi optical patterns, the dense deployments and the multi path diversity warrant designs of localization schemes aimed at exploiting these characteristic. Finally, stepping away from conventional network design, the future generation of networks can be enhanced by Machine Learning. Machine Learning has shown to be a powerful tool in many domains and has increased the autonomy of chatbots and self-driving cars. As we go further in frequency, ML can help design the different layers of the future communication networks to embrace the hardware imperfections, that are more relevant at higher frequencies, and to serve applications in the optimum way in different environments. The thesis is divided into two parts, first we study the challenges in the optimization and features of conventional mm-wave networks, and second, we consider a new way of designing mm-wave communications using Machine Learning. In the first part, we study the challenge of scaling mm-wave networks to dense deployments, and secondly we increase the capabilities of mm-wave networks by proposing a passive localization mechanism that is standard compliant. Currently the beam training mechanisms deployed to establish a directional link result in a large overhead in communications and this prohibits dense deployments. This is specially due to station beam training. While the APs can beam train with multiple stations simultaneously, the stations have to train one at a time. This overhead is even larger in mobile deployments where the devices require constant re-training to maintain communication. To overcome this overhead and allow dense deployments we propose a low-overhead beam training mechanism, SPIDER, that eliminates the station beam training overhead entirely. To this end, stations carry out phase-coherent measurements by switching through multiple receive beam patterns on a time-scale of tens of nanoseconds when receiving a packet preamble to calculate the AoA to the AP and select the best beam pattern for communication. Our results show that our algorithm achieves highly accurate angle estimation used to drive the beam steering decisions and that it reduces the overhead by an order of magnitude compared to IEEE 802.11ad beam training. The second work presented in the first part of this thesis is regarding the use of the network as a sensor in mm-wave frequencies. Joint communication and sensing approaches would allow for an extremely distributed cost-effective sensor network. At higher frequencies, due to the larger bandwidth, the theoretical precision of localization information is higher than at lower frequencies. However, localization algorithms for mm-wave frequencies can not directly build from the lower frequency ones. At lower frequencies, these algorithms leverage information from multiple antennas with separated RF chains, while in mm-wave phased antenna arrays are usually connected to a single RF chain, so the phase information per antenna can not be obtained. However, several properties of mm-wave communications have implications in the design of localization systems. The high path loss attenuation due to the high frequency implies that the line of sight paths (LOS) can be distinguished from the NLOS paths. Additionally, reflections on surfaces have limited scattering. These properties generate channels with sparse multi paths, for which geometrical propagation assumptions become appropriate. Additionally, the need for directional communication due to the high path loss results in the need for mechanisms that enable beam selection, which are the focus of the earlier works. Such beam training mechanisms become relevant for localization in mm-wave frequencies. Since during the beam training the environment is scanned using beam patterns in different directions, these information can be used to detect and track moving reflections. We present POLAR, the first passive mm-wave localization algorithm for mm-wave commercial devices with zero overhead. We use commercial APs whereas the station design is based on a full-bandwidth 802.11ad compatible FPGA-based platform with a phased antenna array. The stations exploit the preamble of the beam training packets of the APs to obtain channel state information measurements for all antenna patterns. With this, we determine distance and angle information for the different multipath components in the environment to passively localize a mobile object. We evaluate our system with multiple APs and a moving robot with a metallic surface. Our results show that is is possible to obtain 6.5cm mean error accuracy and sub-meter accuracy in 100% of the cases performing joint localization and sensing in a mm-wave testbed. In the second part of this thesis, we study a new way to design communications with the concept of AI-native air-interface framework in mm-wave. In this area of research, some works have proposed to learn to communicate using an end-to-end communication approach using deep learning (DL). However, a channel model is needed in order to train these models, and analytical channels were shown to produce a reduced performance in experimental scenarios. In order to extend this to real channels, model-free approaches have been presented in which the model can be learned implicitly from data. Building on this works to obtain an explicit channel model would also allow for transfer learning to speed up the training on time-variant channels. In this thesis, we present a neural network architecture based on Mixture Density Networks to explicitly learn the channel model for end-to-end communications for SISO and MIMO systems. Our results show that it is possible to learn complex stochastic channel models from data with differentiable methods that are compatible with the training of the autoencoder structure. These end-toend learning methods will become increasingly important in future generations of wireless networks, where the system complexity will increase due to a higher number of antennas, higher frequencies, wider bandwidths, and more complex hardware, making ML-based physical layer design highly appealing. Second, we study the scaling properties of end-to-end ML models for large-scale MIMO systems. We conclude that the bottleneck that prevents the scaling of such networks, in the state-of-the-art architectures, is due to the input and output representations. Large MIMO systems constitute highly demanding scenarios in which training would require unmanageable memory resources. We study typical approaches available in DL to reduce the memory allocation and show that a bit-wise architecture for MIMO allows to reduce the memory allocation by several orders of magnitude. Our results show that with the bit-wise architecture it is possible to extend the end-to-end approaches to large systems of antennas. This study is an important first step towards developing practical end-to-end learning algorithms for large scale MIMO systems. So far, end-to-end systems have only been applied in a per symbol manner as training these systems for blocks of symbols is complex as it introduces implicit channel coding. Optimizing the receiver only would allow to reduce the complexity and train in blocks of symbols while helping to deal with imperfections and other non-linearities without having to explicitly correct them. We explore the gains that can be achieved by using an ML based design on the receiver of single carrier mm-wave systems for frequency selective channels, where due to the multi-path and the resulting inter symbol interference block wise treatment is required. Our results show that it is possible to outperform baseline typical receivers in static scenarios. In summary, mm-wave technologies will open the door to the next generation of applications, however, new challenges need to be tackled. We provide new solutions to enable scalable deployments and show that the beam training station overhead can be completely eliminated. We introduce the feature of passive localization and sensing for mm-wave frequencies, and show the high accuracy achievable in an experimental FPGA based testbed. Finally, explore new ways of designing the physical layer to reduce the effect of hardware imperfections in mm-wave and show gains compared to state-of-the-art receivers. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Tobias Mirco Koch.- Secretario: Andrés García Saavedra.- Vocal: Dani Korpi

A main aspect that has characterized the evolution towards the 5th Generation of Mobile Networks (5G) has been the involvement of industrial sectors, or as they are generally known, verticals, in the definition of the requirements that these networks must address with respect to the service provisioning. Thus, this paradigmshift not only considers services that facilitate human communications, but also promotes the creation of a global digital ecosystem in which verticals such as the automotive sector, smart-cities, health-care, or public-safety are key adopters. In addition, this change in the service provisioning model also facilitates the appearance of innovative services (e.g., virtual reality, augmented reality, healthcare, autonomous driving, etc.), which are not only expected to cause an enormous growth in the amount of traffic, but also to demand a high performance from the network. One of the technological advances that has driven this evolution toward 5G has been the adoption of technologies for function virtualization and softwarization supporting the transition from traditional specialized hardware equipment (e.g., IP routers, firewalls or load balancing devices) to versatile software components, which can be deployed in different locations. In particular, 5G embraces NFV, a technology standardized by the ETSI, that enables the automated and agile provision of telecommunication and vertical services in 5G networks, as a composition of virtualized components, commonly referred to as Virtualized Network Functions (VNFs). However, under the temporal context in which the beginning of this thesis is situated (i.e., in 2017), NFV was starting to receive a great interest from the industry and research community, and there were just a few open source initiatives aiming to implement the standard. Due to this, an additional effort was needed to understand, by means of existent implementations at that time, the implications and challenges of applying the NFV standards in practical situations. In this context, the first part of this thesis is devoted to creating a platform capable of enabling complex, close to reality, experimentation scenarios across a distributed set of NFV and vertical infrastructures, which can be made available by different stakeholders at different geographic locations. Thus, consolidating important aspects of the standard, as well as identifying new necessary specifications. In accordance to these specifications, the next part of this thesis is intended to address the lack of flexibility on the 5G networks to support reliable service operations in environments and situations where there are obvious resource constraints. For instance, in (i) remote areas where 5G radio access network coverage is insufficient or non-existent; (ii) emergency situations (e.g., natural disasters), where the network infrastructure may fail or provide deficient service; or (iii) situations where there are occasional high, unexpected or predictable, service demands such as in the case of mass events. To this purpose, this thesis explores the potential benefits of creating an NFV system based on Unmanned Aerial Vehicles (UAVs) to exploit the inherent capabilities of this aerial devices, and extend the programmable substrate of 5G networks beyond the network access segments of telecommunications operators. Then, the thesis analyzes the potential benefits of using the UAV-based system in the deployment of services included within the context of different vertical sectors. Firstly, this part analyzes possible synergies between NFV, UAVs, and vertical services from a practical perspective, presenting the creation of a multi-site testbed at national scale to support prototyping, and experimentation activities. This testbed builds on the NFV system based on UAVs, and on themechanisms related to the orchestration of telecommunications and vertical services within a multi-site NFV ecosystem, previously discussed in the context of this thesis. Moreover, this testbed allows to realize the practical validation related to the capability of the UAVs-based system to support vertical services. This validation is performed then with the definition of a use case involving smart-farming vertical, instantiating a precision agriculture service over the UAVs on a remote site. Subsequently, following this line, this thesis explores the interoperation of the UAV-based system with other NFV infrastructures, with the aim of supporting the deployment of telecommunications and/or vertical services in resource-constrained situations. In particular, this part considers a situation related to the public-safety vertical, where the system collaborates with an NFV infrastructure composed of a fleet of vehicles to assist a response teamin the management of an emergency. To conclude, this thesis addresses an additional key challenge to support the provision of telecommunications and vertical services in 5G: the provision of adequate mechanisms to enable the exchange of data traffic between VNFs, which may be located in different 5G domains. This is not only interesting from the general point of view of the NFV ecosystems, but this type of communications particularized for the UAVs-based system, could allow this system to collaborate with other NFV infrastructures e.g., cloud/edge platforms) to support the proper operations of more elaborated services. In this context, this last part of the thesis presents a novel solution to support secure link-layer connectivity for virtualized functions in multi-site NFV ecosystems. Thus, providing an appropriate mechanism to enable the exchange of data traffic among VNFs that are located in different NFV domains. As part of this work, the thesis explores the use of a SDN framework to evolve such a solution, offering an inter-domain connectivity orchestration service, which intends to support the automated and on-demand provisioning and configuration of virtual networks between different NFV domains. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Ignacio Soto Campos.- Secretario: Marco Gramaglia.- Vocal: Luis Miguel Contreras Murillo

Until today, Unmanned Aerial Vehicle (UAV) operations only include a single aerial vehicle (in most cases) that performs reconnaissance missions by sending telemetry captured by different onboarded sensors (e.g., video, temperature, air quality) to the Ground Control Station (GCS). Single-UAV applications, despite their apparent simplicity, are used in many different and significant fields (e.g., surveillance of livestock, monitoring of power lines, traffic monitoring, rescue). Many applications of UAV swarms have already been seen. Still, they are usually stunts and exhibitions with no actual functionality. Recent research trends are founded on multiple UAVs operating collaborative implementing more complex services, and generally integrated into the urban environment. It would lead to new scenarios that are not yet adequately deployed (e.g., package delivery, monitoring of sports events or crowds such as concerts or demonstrations, increasing coverage, support to emergency services in cities (fire, police, emergency)). However, several challenges must be faced before integrating these applications into our everyday lives. The central objective of the thesis is to contribute to some of the significant challenges identified in the UAV communications services sector. In the first place, this thesis contributes with an emulation solution for validating environments with connected UAVs, including different use cases and verticals. Additionally, it contributes to communications solutions in complex connectivity environments based on experimentation where the Fifth Generation of cellular network technology (5G) softwarization technologies are integrated into the UAV ecosystem. In the last place, this thesis contributes to the proposal of new solutions to solve some limitations, such as the high energy consumption in combination with UAVs’ limited flight autonomy or the complexity of traffic management and the establishment of the network infrastructure in such volatile environments. This doctoral thesis has been carried out in the framework of the European and National projects listed below: 5GCity: Adaptive Management of 5G Services to Support Critical Events in Cities (Grant agreement TEC2016-76795-C6-3-R, Spanish Ministry of Economy and Competitiveness). TRUE5G: Towards zeRo toUch nEtwork and services for beyond 5G (Grant agreement PID2019-108713RB681). H2020 5GRANGE: Remote area Access Network for 5th GEneration (Grant agreement 777137). H2020 Labyrinth: Unmanned Traffic Management 4D path planning technologies for drone swarm to enhance safety and security in transport (Grant agreement 861696). This work has been supported by IMDEA Networks Institute. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: José Ignacio Moreno Novella.- Secretario: María Carmen Guerrero López.- Vocal: Elisa Rojas Sánchez

Mención Internacional en el título de doctor Localization has been mainly an optional feature of cellular networks since they have been designed for communication and many location-based services have used GPS to provide more functionalities like navigation, rescue and many more. GPS gets outstanding accuracy in outdoor environments, but its performance drastically degrades in indoor settings since the GPS signals hardly go through walls. However, most human activities are concentrated in indoor environments and location-based services cannot be carried out successfully by GPS. To overcome this limitation, wireless protocols are appealing to fulfill indoor localization requirements while communication is ongoing. Operators, chipset vendors and application developers are paying attention to exploiting location information to provide new applications like augmented reality and indoor navigation. Moreover, localization can be used for network optimization and researchers are actively investigating it. For instance, intelligent handover can exploit location information to guess which Access Point (AP) is the most suitable one before doing the handover. In addition, a range of applications can exploit it as well such as Multiple-Input Multiple-Output beamforming, millimeter-wave beam alignment, etc. For the last decade, sensing has been appealing to not only provide location information but also context awareness. This enables human activity and event recognition, vital sign monitoring, user identification, mapping, imaging, etc. To ensure the good performance of these applications, accurate and ubiquitous positioning is needed. To this end, 5G and the newest Wi-Fi protocols, IEEE 802.11ac and 802.11ax, are becoming the key technologies to provide outstanding indoor localization since they incorporate larger array configurations and wider channel bandwidths than previous wireless protocols. Researchers have made a great effort to provide indoor localization and decimeter level of accuracy has been achieved. However, this outstanding performance has been evaluated using a great number of APs and assuming that every AP has a clear Line-Of- Sight (LOS) to the device. However, typical indoor wireless deployments tend to have sparse AP densities since they are optimized for coverage and not for localization. For instance, a Wi-Fi infrastructure usually contains one AP per room and a 5G deployment tends to have a limited number of AP as well. Moreover, indoor environments are generally rich in multipath components that interfere with the estimation of the direct path. This is particularly challenging in Non-Line-Of-Sight (NLOS) settings as obstacles can block the direct path and a system might detect an NLOS path and not the obstructed LOS path. As a result, the performances of state-of-the-art localization schemes drastically degrade their accuracy in realistic deployments. A localization algorithm that copes well with NLOS settings and wireless deployments with sparse AP densities is needed for precise and pervasive localization. Also, implementing and testing it in cutting edge devices is crucial to exploit the improved hardware features of the newest wireless protocols. Therefore, this thesis aims at providing a framework for accurate localization even in challenging scenarios. Sensing research shares methodologies with localization since sensing applications require extracting location information from NLOS paths as localization does from the direct path. Hence, this thesis also aims at exploring how the proposed localization framework can be used for sensing applications. We start delving into wireless localization by exploring what an LTE localization system can achieve. This is particularly beneficial since 5G and LTE will coexist for a while until 5G provides ubiquitous coverage. Therefore, LTE needs to fulfill the localization requirements for a range of applications if 5G is not available. To this end, we implement and evaluate an LTE localization system for a single AP using software-defined radios. We observe that LTE achieves a median error of 2 m in LOS cases. However, the LTE performance drastically degrades to 4.6 m of median error in NLOS settings. These results point out that LTE provides a positioning accuracy that complies with a great number of location-based services in LOS. Nevertheless, applications that demand ubiquitous localization may not be correctly carried out in NLOS settings. To tackle the NLOS issue, we implement UbiLocate, a Wi-Fi location system that copes well with common AP deployment densities and works ubiquitously, i.e., without excessive degradation under NLOS. UbiLocate demonstrates that meter-level median accuracy NLOS localization is possible through (i) an innovative angle estimator based on a Nelder-Mead search, (ii) a fine-grained time of flight ranging system with nanosecond resolution, and (iii) the accuracy improvements brought about by the increase in bandwidth and number of antennas of IEEE 802.11ac. In combination, they provide superior resolvability of multipath components, significantly improving location accuracy over prior work. We implement our location system on off-the-shelf 802.11ac devices. Our experimental evaluation shows an overall improvement of the localization performance by a factor of 2-3. The latest generation of Wi-Fi standards, IEEE 802.11ax, brings new hardware capabilities that improve the performance of localization and sensing systems. In particular, the 160MHz of channel bandwidth and the four times denser spectrum significantly improve the resolvability of the multipath components compared to its predecessor, IEEE 802.11ac. We present the first tool to collect the most accurate CSI ever from off-the-shelf devices. To further validate the platform, we carry out a preliminary measurement campaign to compare the localization accuracy of IEEE 802.11ax with 802.11ac. Our results show that, as expected, IEEE 802.11ax provides superior performance improving the accuracy by a factor of 1.75 for LOS and NLOS settings. Sensing research goes beyond localization since it aims at providing context awareness. We explore the integration of the proposed multipath decomposition algorithm as well as the testbed for sensing applications. In particular, we tackle human respiration rate estimation since it is appealing as it does not require any specialized hardware. Our results show that an accurate respiration rate estimation is possible by decomposing the channel. In summary, location-based services demand accurate and ubiquitous localization. However, the state-of-the-art localization systems do not cope well with realistic wireless deployments and their positioning performances drastically degrade in these environments. Hence, we provide a localization framework that copes well with realistic wireless deployments and with NLOS settings. We conclude that resolving accurately the multipath components enables pervasive and precise localization. In addition, sensing enables new applications that are helpful in many issues since it provides not only location but also context awareness. Hence, we show that algorithms and testbeds that are designed for localization can be also utilized for sensing applications by tackling respiration rate estimation. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Vincenzo Mancuso.- Secretaria: Andrea Conti.- Vocal: Paolo Casari

With the rapidly growing market for smartphones and user’s confidence for immediate access to high-quality multimedia content, the delivery of video over wireless networks has become a big challenge. It makes it challenging to accommodate end-users with flawless quality of service. The growth of the smartphone market goes hand in hand with the development of the Internet, in which current transport protocols are being re-evaluated to deal with traffic growth. QUIC and WebRTC are new and evolving standards. The latter is a unique and evolving standard explicitly developed to meet this demand and enable a high-quality experience for mobile users of real-time communication services. QUIC has been designed to reduce Web latency, integrate security features, and allow a highquality experience for mobile users. Thus, the need to evaluate the performance of these rising protocols in a non-systematic environment is essential to understand the behavior of the network and provide the end user with a better multimedia delivery service. Since most of the work in the research community is conducted in a controlled environment, we leverage the MONROE platform to investigate the performance of QUIC and WebRTC in real cellular networks using static and mobile nodes. During this Thesis, we conduct measurements ofWebRTC and QUIC while making their data-sets public to the interested experimenter. Building such data-sets is very welcomed with the research community, opening doors to applying data science to network data-sets. The development part of the experiments involves building Docker containers that act as QUIC and WebRTC clients. These containers are publicly available to be used candidly or within the MONROE platform. These key contributions span from Chapter 4 to Chapter 5 presented in Part II of the Thesis. We exploit data collection from MONROE to apply data science over network data-sets, which will help identify networking problems shifting the Thesis focus from performance evaluation to a data science problem. Indeed, the second part of the Thesis focuses on interpretable data science. Identifying network problems leveraging Machine Learning (ML) has gained much visibility in the past few years, resulting in dramatically improved cellular network services. However, critical tasks like troubleshooting cellular networks are still performed manually by experts who monitor the network around the clock. In this context, this Thesis contributes by proposing the use of simple interpretable ML algorithms, moving away from the current trend of high-accuracy ML algorithms (e.g., deep learning) that do not allow interpretation (and hence understanding) of their outcome. We prefer having lower accuracy since we consider it interesting (anomalous) the scenarios misclassified by the ML algorithms, and we do not want to miss them by overfitting. To this aim, we present CIAN (from Causality Inference of Anomalies in Networks), a practical and interpretable ML methodology, which we implement in the form of a software tool named TTrees (from Troubleshooting Trees) and compare it to a supervised counterpart, named STress (from Supervised Trees). Both methodologies require small volumes of data and are quick at training. Our experiments using real data from operational commercial mobile networks e.g., sampled with MONROE probes, show that STrees and CIAN can automatically identify and accurately classify network anomalies—e.g., cases for which a low network performance is not justified by operational conditions—training with just a few hundreds of data samples, hence enabling precise troubleshooting actions. Most importantly, our experiments show that a fully automated unsupervised approach is viable and efficient. In Part III of the Thesis which includes Chapter 6 and 7. In conclusion, in this Thesis, we go through a data-driven networking roller coaster, from performance evaluating upcoming network protocols in real mobile networks to building methodologies that help identify and classify the root cause of networking problems, emphasizing the fact that these methodologies are easy to implement and can be deployed in production environments. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Multimedia y Comunicaciones por la Universidad Carlos III de Madrid y la Universidad Rey Juan Carlos Presidente: Matteo Sereno.- Secretario: Antonio de la Oliva Delgado.- Vocal: Raquel Barco Moreno

Mención Internacional en el título de doctor Reinforcement Learning (RL), in which an agent is trained to make the most favourable decisions in the long run, is an established technique in artificial intelligence. Its popularity has increased in the recent past, largely due to the development of deep neural networks spawning deep reinforcement learning algorithms such as Deep Q-Learning. The latter have been used to solve previously insurmountable problems, such as playing the famed game of “Go” that previous algorithms could not. Many such problems suffer the curse of dimensionality, in which the sheer number of possible states is so overwhelming that it is impractical to explore every possible option. While these recent techniques have been successful, they may not be strictly necessary or practical for some applications such as cloud provisioning. In these situations, the action space is not as vast and workload data required to train such systems is not as widely shared, as it is considered commercialy sensitive by the Application Service Provider (ASP). Given that provisioning decisions evolve over time in sympathy to incident workloads, they fit into the sequential decision process problem that legacy RL was designed to solve. However because of the high correlation of time series data, states are not independent of each other and the legacy Markov Decision Processes (MDPs) have to be cleverly adapted to create robust provisioning algorithms. As the first contribution of this thesis, we exploit the knowledge of both the application and configuration to create an adaptive provisioning system leveraging stationary Markov distributions. We then develop algorithms that, with neither application nor configuration knowledge, solve the underlying Markov Decision Process (MDP) to create provisioning systems. Our Q-Learning algorithms factor in the correlation between states and the consequent transitions between them to create provisioning systems that do not only adapt to workloads, but can also exploit similarities between them, thereby reducing the retraining overhead. Our algorithms also exhibit convergence in fewer learning steps given that we restructure the state and action spaces to avoid the curse of dimensionality without the need for the function approximation approach taken by deep Q-Learning systems. A crucial use-case of future networks will be the support of low-latency applications involving highly mobile users. With these in mind, the European Telecommunications Standards Institute (ETSI) has proposed the Multi-access Edge Computing (MEC) architecture, in which computing capabilities can be located close to the network edge, where the data is generated. Provisioning for such applications therefore entails migrating them to the most suitable location on the network edge as the users move. In this thesis, we also tackle this type of provisioning by considering vehicle platooning or Cooperative Adaptive Cruise Control (CACC) on the edge. We show that our Q-Learning algorithm can be adapted to minimize the number of migrations required to effectively run such an application on MEC hosts, which may also be subject to traffic from other competing applications. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Antonio Fernández Anta.- Secretario: Diego Perino.- Vocal: Ilenia Tinnirello

3 January 2009, the first version of the Bitcoin software was released. That date has been the really first time that blockchain technology has been revealed to the world. A blockchain is an immutable distributed ledger that records validated transactions permanently employing cryptography and a distributed consensus algorithms. All the information is stored and validated by all the nodes composing the peer-to-peer network without the need of a trusted third party. Bitcoin is the first application of this technology and it allows e-payments to be easily performed between two parties with no need of intermediaries. Since then the hype about the technology has led several advances, exceeding the pure financial sphere and attempting to solve disparate problems, rather than digital payments only. As example, Ethereum with its introduction of smart contracts, has been the first application of the so called 2.0 generation of blockchain technology. A smart contract is a self-executing program, stored in the blockchain, with the terms of the agreement between interested parties hard-coded in the contract definition. Smart contract allow trusted transactions and agreements to be performed between distrust parties without the need for any intermediary or external enforcement mechanism. Therefore, with the introduction of smart contracts, the blockchain paradigm can be extended to the automation of complex resource manipulation and transference procedures in a transparent and trustable manner, by means of the specification of these new types of contracts. Blockchain and smart contracts technology is being adopted in a vast number of scenarios, including finances, Internet of Things, health care, energy, education, and more. As with any novel technology, there are many open questions about its usage and potential. For this reason, we believe that practical experimentation is in order to have hands-on experiences. So we present this research work on the application of the blockchain technology for the management of internet resources. The first study we present in this thesis is the design of a blockchain based application to build a decentralised IP address registry. Empowering the blockchain technology we propose to change by design the centralized structure of the current system to manage the global pool of IP addresses and the centralized and hierarchical model that is implemented in the Resource Public Key Infrastructure (RPKI) that makes lower layers in the hierarchy susceptible to errors and abuses from entities placed in higher layers. Hence we present the design of InBlock, a Distributed Autonomous Organization (DAO) that provides decentralized management of IP addresses. The InBlock automates the process of assigning Internet resources to the user complying with the "law" written in its smart contracts. InBlock also fulfills the same objectives as the current IP address allocation organizations, i.e., uniqueness, fairness, conservation, aggregation, registration and minimized overhead. InBlock is implemented as a set of blockchain’s smart contracts in Ethereum and it implements all the functions needed for the management of a global pool of addresses without any human intervention. Moreover InBlock embeds an alternative trust model to the hierarchical one currently implemented by the RPKI. In this thesis we present two Proof of Concept (PoC) implementation of InBlock: InBlock6 and InBlock4. InBlock6 implements the InBlock design and is centred on the management of the IPv6 address space, that compared to the IPv4 address space, has way more free resources that can be assigned. InBlock4 inherits its functionalities but for the IPv4 address space and provides an alternative framework to register living resources (e.g. already assigned resources) into the blockchain to enable the decentralised route origin validation. We present the implementation and evaluation of both the PoC for the Ethereum blockchain and we quantify their performance. The second study we present in this thesis is on the design and a PoC implementation of the Internet Routing Blockchain (IRB), an implementation of the Internet Routing Registry (IRR) functionality within Hyperledger Fabric (HF). The IRR is a distributed routing database that provides a mechanism for validating the contents of Border Gateway Protocol (BGP) announcement messages and mapping an origin Autonomous System (AS) number to a list of networks [6]. The IRB relies on a permissioned blockchain technology that is inherently distributed, allows to preserve the decentralised nature of the IRR, overcomes the centralized governance model limitation of current used mechanism employing a consortium based model, provides consistency and information stall-ness prevention and offers a simple declaration syntax for the policy definition of ASes relationship. As the final contribution of this thesis we present a study about route leaks prevention and the utilization of the information stored in the IRR. A route leak is defined as the propagation of a route beyond its intended scope. Those events have historically caused a consistent number of incidents resulting in Internet disconnections and disservices that generates money loss. In this research work we present the design and the performance evaluation of Autonomous System Internet Registry Inference for path Authorization (ASIRIA), a mechanism for detecting leaked routes and leakage events that uses AS relationship information inferred from the IRR. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Diego Rafael López García.- Secretario: Francisco Valera Pintor.- Vocal: Anna Maria Mandalari

Mención Internacional en el título de doctor Las infraestructuras de red actuales soportan una variedad diversa de servicios como video bajo demanda, video conferencias, redes sociales, sistemas de educación, o servicios de almacenamiento de fotografías. Gran parte de la población mundial ha comenzado a utilizar estos servicios, y los utilizan diariamente. Proveedores de Cloud y operadores de infraestructuras de red albergan el tráfico de red generado por estos servicios, y sus tareas de gestión no solo implican realizar el enrutamiento del tráfico, sino también el procesado del tráfico de servicios de red. Tradicionalmente, el procesado del tráfico ha sido realizado mediante aplicaciones/ programas desplegados en servidores que estaban dedicados en exclusiva a tareas concretas como la inspección de paquetes. Sin embargo, en los últimos anos los servicios de red se han virtualizado y esto ha dado lugar al paradigma de virtualización de funciones de red (Network Function Virtualization (NFV) siguiendo las siglas en ingles), en el que las funciones de red de un servicio se ejecutan en contenedores o máquinas virtuales desacopladas de la infraestructura hardware. Como resultado, el procesado de tráfico se ha ido haciendo más flexible gracias al laxo acople del software y hardware, y a la posibilidad de compartir funciones de red típicas, como firewalls, entre los distintos servicios de red. NFV facilita la automatización de operaciones de red, ya que tareas como el escalado, o la migración son típicamente llevadas a cabo mediante un conjunto de comandos previamente definidos por la tecnología de virtualización pertinente, bien mediante contenedores o máquinas virtuales. De todos modos, sigue siendo necesario decidir el en rutamiento y procesado del tráfico de cada servicio de red. En otras palabras, que servidores tienen que encargarse del procesado del tráfico, y que enlaces de la red tienen que utilizarse para que las peticiones de los usuarios lleguen a los servidores finales, es decir, el conocido como embedding problem. Bajo el paraguas del paradigma NFV, a este problema se le conoce en inglés como Virtual Network Embedding (VNE), y esta tesis utiliza el termino “NFV orchestration algorithm” para referirse a los algoritmos que resuelven este problema. El problema del VNE es NP-hard, lo cual significa que que es imposible encontrar una solución optima en un tiempo polinómico, independientemente del tamaño de la red. Como consecuencia, la comunidad investigadora y de telecomunicaciones utilizan heurísticos que encuentran soluciones de manera más rápida que productos para la resolución de problemas de optimización. Tradicionalmente, los “NFV orchestration algorithms” han intentado minimizar los costes de despliegue derivados de las soluciones asociadas. Por ejemplo, estos algoritmos intentan no consumir el ancho de banda de la red, y usar rutas cortas para no utilizar tantos recursos. Además, una tendencia reciente ha llevado a la comunidad investigadora a utilizar algoritmos que minimizan el consumo energético de los servicios desplegados, bien mediante la elección de dispositivos con un consumo energético más eficiente, o mediante el apagado de dispositivos de red en desuso. Típicamente, las restricciones de los problemas de VNE se han resumido en un conjunto de restricciones asociadas al uso de recursos y consumo energético, y las soluciones se diferenciaban por la función objetivo utilizada. Pero eso era antes de la 5a generación de redes móviles (5G) se considerase en el problema de VNE. Con la aparición del 5G, nuevos servicios de red y casos de uso entraron en escena. Los estándares hablaban de comunicaciones ultra rápidas y fiables (Ultra-Reliable and Low Latency Communications (URLLC) usando las siglas en inglés) con latencias por debajo de unos pocos milisegundos y fiabilidades del 99.999%, una banda ancha mejorada (enhanced Mobile Broadband (eMBB) usando las siglas en inglés) con notorios incrementos en el flujo de datos, e incluso la consideración de comunicaciones masivas entre maquinas (Massive Machine-Type Communications (mMTC) usando las siglas en inglés) entre dispositivos IoT. Es más, paradigmas como edge y fog computing se incorporaron a la tecnología 5G, e introducían la idea de tener dispositivos de computo más cercanos al usuario final. Como resultado, el problema del VNE tenía que incorporar los nuevos requisitos como restricciones a tener en cuenta, y toda solución debía satisfacer bajas latencias, alta fiabilidad, y mayores tasas de transmisión. Esta tesis estudia el problema des VNE, y propone algunos heurísticos que lidian con las restricciones asociadas a servicios 5G en escenarios edge y fog, es decir, las soluciones propuestas se encargan de asignar funciones virtuales de red a servidores, y deciden el enrutamiento del trafico en las infraestructuras 5G con dispositivos edge y fog. Para evaluar el rendimiento de las soluciones propuestas, esta tesis estudia en primer lugar la generación de grafos que representan redes 5G. Los mecanismos propuestos para la generación de grafos sirven para representar distintos escenarios 5G. En particular, escenarios de federación en los que varios dominios comparten recursos entre ellos. Los grafos generados también representan servidores en el edge, así como dispositivos fog con una batería limitada. Además, estos grafos tienen en cuenta los requisitos de estándares, y la demanda que se espera en las redes 5G. La generación de grafos propuesta sirve para representar escenarios federación en los que varios dominios comparten recursos entre ellos, y redes 5G con servidores edge, así como dispositivos fog estáticos o móviles con una batería limitada. Los grafos generados para infraestructuras 5G tienen en cuenta los requisitos de estándares, y la demanda de red que se espera en las redes 5G. Además, los grafos son diferentes en función de la densidad de población, y el área de estudio, es decir, si es una zona industrial, una autopista, o una zona urbana. Tras detallar la generación de grafos que representan redes 5G, esta tesis propone algoritmos de orquestación NFV para resolver con el problema del VNE. Primero, se centra en escenarios federados en los que los servicios de red se tienen que asignar no solo a la infraestructura de un dominio, sino a los recursos compartidos en la federación de dominios. Dos problemas diferentes han sido estudiados, uno es el problema del VNE propiamente dicho sobre una infraestructura federada, y el otro es la delegación de servicios de red. Es decir, si un servicio de red se debe desplegar localmente en un dominio, o en los recursos compartidos por la federación de dominios; a sabiendas de que el último caso supone el pago de cuotas por parte del dominio local a cambio del despliegue del servicio de red. En segundo lugar, esta tesis propone OKpi, un algoritmo de orquestación NFV para conseguir la calidad de servicio de las distintas slices de las redes 5G. Conceptualmente, el slicing consiste en partir la red de modo que cada servicio de red sea tratado de modo diferente dependiendo del trozo al que pertenezca. Por ejemplo, una slice de eHealth reservara los recursos de red necesarios para conseguir bajas latencias en servicios como operaciones quirúrgicas realizadas de manera remota. Cada trozo (slice) está destinado a unos servicios específicos con unos requisitos muy concretos, como alta fiabilidad, restricciones de localización, o latencias de un milisegundo. OKpi es un algoritmo de orquestación NFV que consigue satisfacer los requisitos de servicios de red en los distintos trozos, o slices de la red. Tras presentar OKpi, la tesis resuelve el problema del VNE en redes 5G con dispositivos fog estáticos y móviles. El algoritmo de orquestación NFV presentado tiene en cuenta las limitaciones de recursos de computo de los dispositivos fog, además de los problemas de falta de cobertura derivados de la movilidad de los dispositivos. Para concluir, esta tesis estudia el escalado de servicios vehiculares Vehicle-to-Network (V2N), que requieren de bajas latencias para servicios como la prevención de choques, avisos de posibles riesgos, y conducción remota. Para estos servicios, los atascos y congestiones en la carretera pueden causar el incumplimiento de los requisitos de latencia. Por tanto, es necesario anticiparse a esas circunstancias usando técnicas de series temporales que permiten saber el tráfico inminente en los siguientes minutos u horas, para así poder escalar el servicio V2N adecuadamente. Current network infrastructures handle a diverse range of network services such as video on demand services, video-conferences, social networks, educational systems, or photo storage services. These services have been embraced by a significant amount of the world population, and are used on a daily basis. Cloud providers and Network operators’ infrastructures accommodate the traffic rates that the aforementioned services generate, and their management tasks do not only involve the traffic steering, but also the processing of the network services’ traffic. Traditionally, the traffic processing has been assessed via applications/programs deployed on servers that were exclusively dedicated to a specific task as packet inspection. However, in recent years network services have stated to be virtualized and this has led to the Network Function Virtualization (Network Function Virtualization (NFV)) paradigm, in which the network functions of a service run on containers or virtual machines that are decoupled from the hardware infrastructure. As a result, the traffic processing has become more flexible because of the loose coupling between software and hardware, and the possibility of sharing common network functions, as firewalls, across multiple network services. NFV eases the automation of network operations, since scaling and migrations tasks are typically performed by a set of commands predefined by the virtualization technology, either containers or virtual machines. However, it is still necessary to decide the traffic steering and processing of every network service. In other words, which servers will hold the traffic processing, and which are the network links to be traversed so the users’ requests reach the final servers, i.e., the network embedding problem. Under the umbrella of NFV, this problem is known as Virtual Network Embedding (VNE), and this thesis refers as “NFV orchestration algorithms” to those algorithms solving such a problem. The VNE problem is a NP-hard, meaning that it is impossible to find optimal solutions in polynomial time, no matter the network size. As a consequence, the research and telecommunications community rely on heuristics that find solutions quicker than a commodity optimization solver. Traditionally, NFV orchestration algorithms have tried to minimize the deployment costs derived from their solutions. For example, they try to not exhaust the network bandwidth, and use short paths to use less network resources. Additionally, a recent tendency led the research community towards algorithms that minimize the energy consumption of the deployed services, either by selecting more energy efficient devices or by turning off those network devices that remained unused. VNE problem constraints were typically summarized in a set of resources/energy constraints, and the solutions differed on which objectives functions were aimed for. But that was before 5th generation of mobile networks (5G) were considered in the VNE problem. With the appearance of 5G, new network services and use cases started to emerge. The standards talked about Ultra Reliable Low Latency Communication (Ultra-Reliable and Low Latency Communications (URLLC)) with latencies below few milliseconds and 99.999% reliability, an enhanced mobile broadband (enhanced Mobile Broadband (eMBB)) with significant data rate increases, and even the consideration of massive machine-type communications (Massive Machine-Type Communications (mMTC)) among Internet of Things (IoT) devices. Moreover, paradigms such as edge and fog computing blended with the 5G technology to introduce the idea of having computing devices closer to the end users. As a result, the VNE problem had to incorporate the new requirements as constraints to be taken into account, and every solution should either satisfy low latencies, high reliability, or larger data rates. This thesis studies the VNE problem, and proposes some heuristics tackling the constraints related to 5G services in Edge and fog scenarios, that is, the proposed solutions assess the assignment of Virtual Network Functions to resources, and the traffic steering across 5G infrastructures that have Edge and Fog devices. To evaluate the performance of the proposed solutions, the thesis studies first the generation of graphs that represent 5G networks. The proposed mechanisms to generate graphs serve to represent diverse 5G scenarios. In particular federation scenarios in which several domains share resources among themselves. The generated graphs also represent edge servers, so as fog devices with limited battery capacity. Additionally, these graphs take into account the standard requirements, and the expected demand for 5G networks. Moreover, the graphs differ depending on the density of population, and the area of study, i.e., whether it is an industrial area, a highway, or an urban area. After detailing the generation of graphs representing the 5G networks, this thesis proposes several NFV orchestration algorithms to tackle the VNE problem. First, it focuses on federation scenarios in which network services should be assigned not only to a single domain infrastructure, but also to the shared resources of the federation of domains. Two different problems are studied, one being the VNE itself over a federated infrastructure, and the other the delegation of network services. That is, whether a network service should be deployed in a local domain, or in the pool of resources of the federation domain; knowing that the latter charges the local domain for hosting the network service. Second, the thesis proposes OKpi, a NFV orchestration algorithm to meet 5G network slices quality of service. Conceptually, network slicing consists in splitting the network so network services are treated differently based on the slice they belong to. For example, an eHealth network slice will allocate the network resources necessary to meet low latencies for network services such as remote surgery. Each network slice is devoted to specific services with very concrete requirements, as high reliability, location constraints, or 1ms latencies. OKpi is a NFV orchestration algorithm that meets the network service requirements among different slices. It is based on a multi-constrained shortest path heuristic, and its solutions satisfy latency, reliability, and location constraints. After presenting OKpi, the thesis tackles the VNE problem in 5G networks with static/moving fog devices. The presented NFV orchestration algorithm takes into account the limited computing resources of fog devices, as well as the out-of-coverage problems derived from the devices’ mobility. To conclude, this thesis studies the scaling of Vehicle-to-Network (V2N) services, which require low latencies for network services as collision avoidance, hazard warning, and remote driving. For these services, the presence of traffic jams, or high vehicular traffic congestion lead to the violation of latency requirements. Hence, it is necessary to anticipate to such circumstances by using time-series techniques that allow to derive the incoming vehicular traffic flow in the next minutes or hours, so as to scale the V2N service accordingly. The 5G Exchange (5GEx) project (2015-2018) was an EU-funded project (H2020-ICT-2014-2 grant agreement 671636). The 5G-TRANSFORMER project (2017-2019) is an EU-funded project (H2020-ICT-2016-2 grant agreement 761536). The 5G-CORAL project (2017-2019) is an EU-Taiwan project (H2020-ICT-2016-2 grant agreement 761586). Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Ioannis Stavrakakis.- Secretario: Pablo Serrano Yáñez-Mingot.- Vocal: Paul Horatiu Patras

El ecosistema web es enorme y, en general, se sustenta principalmente en un atributo intangible que sostiene la mayoría de los servicios gratuitos: la explotación de la información personal del usuario. A lo largo de los años, la preocupación por la forma en que los servicios utilizan los datos personales ha aumentado y atraído la atención de los medios de comunicación, gobiernos, reguladores y también de los usuarios. Esta recogida de información personal es hoy en día la principal fuente de ingresos en Internet. Además, por si fuera poco, la publicidad online es la pieza que lo sustenta todo. Sin la existencia de datos personales en comunión con la publicidad online, Internet probablemente no sería el gigante que hoy conocemos. La publicidad online es un ecosistema muy complejo en el que participan múltiples actores. Es el motor principal que genera ingresos en la red, y en pocos años ha evolucionado hasta llegar a miles de millones de usuarios en todo el mundo. Mientras navegan, los usuarios generan datos muy valiosos sobre sí mismos que los anunciantes utilizan después para ofrecerles productos relevantes en los que podrían estar interesados. Se trata de un enfoque bidireccional, ya que los anunciantes pagan a intermediarios para que muestren anuncios al público que, en principio, está más interesado. Sin embargo, este comercio, intercambio y tratamiento de datos personales, además de abrir nuevas vías de publicidad, exponen la privacidad de los usuarios. Esta incesante recopilación y comercialización de la información personal suele quedar tras un muro opaco, donde el usuario generalmente desconoce para qué se utilizan sus datos. Las iniciativas de privacidad y transparencia se han incrementado a lo largo de los años para empoderar al usuario en este negocio que mueve miles de millones de dólares en ingresos. No en vano, tras varios escándalos, como el de Facebook Cambridge Analytica, las empresas y los reguladores se han unido para crear transparencia y proteger a los usuarios de las malas prácticas derivadas del uso de su información personal. Por ejemplo, el Reglamento General de Protección de Datos, es el ejemplo más prometedor de regulación, que afecta a todos los estados miembros de la Unión Europea, abogando por la protección de los usuarios. El contenido de esta tesis tomará como referencia esta legislación. Por todo ello, el propósito de esta tesis consiste en aportar herramientas y metodologías que pongan de manifiesto usos inapropiados de datos personales por las grandes compañías del ecosistema publicitario online, y cree transparencia entre los usuarios, proporcionando, a su vez, soluciones para que se protejan. Así pues, el contenido de esta tesis ofrece diseño, análisis e implementación de metodologías que miden el impacto social y económico de la información personal online en los servicios extensivos de Internet. Principalmente, se centra en Facebook, una de las mayores redes sociales y servicios en la web, que cuenta con más de 2,8B de usuarios en todo el mundo y generó unos ingresos solo en publicidad online de más de 84 mil millones de dólares en el año 2020. En primer lugar, esta tesis presenta una solución, en forma de extensión del navegador llamada FDVT (Data Valuation Tool for Facebook users), para proporcionar a los usuarios una estimación personalizada y en tiempo real del dinero que están generando para Facebook. Analizando el número de anuncios e interacciones en una sesión, el usuario obtiene información sobre su valor dentro de esta red social. La extensión del navegador ha tenido una importante repercusión y adopción tanto por parte de los usuarios, instalándose más de 10k veces desde su lanzamiento público en octubre de 2016, como de los medios de comunicación, apareciendo en más de 100 medios. En segundo lugar, el estudio e investigación de los posibles riesgos asociados al tratamiento de los datos de los usuarios debe seguir también a la creación de este tipo de soluciones. En este contexto, esta tesis descubre y desvela resultados impactantes sobre el uso de la información personal: (i) cuantifica el número de usuarios afectados por el uso de atributos sensibles utilizados para la publicidad en Facebook, utilizando como referencia la definición de datos sensibles del Reglamento General de Protección de Datos. Esta tesis se basa en el uso de Procesamiento de Lenguaje Natural para identificar los atributos sensibles, y posteriormente utiliza el la plataforma de creación de anuncios de Facebook para recuperar el número de usuarios asignados con esta información sensible. Dos tercios de los usuarios de Facebook se ven afectados por el uso de datos personales sensibles que se les atribuyen. Además, la legislación parece no tener efecto en este uso de atributos sensibles por parte de Facebook, y presenta graves riesgos para los usuarios. (ii) Se modela cuál es el número de atributos que no identifican a priori personalmente al usuario y que aun así son suficientes para identificar de forma única a un individuo sobre una base de datos de miles de millones de usuarios, y se demuestra que llegar a un solo usuario es plausible incluso sin conocer datos que lo identifiquen personalmente de ellos mismos. Los resultados demuestran que 22 intereses al azar de un usuario son suficientes para identificarlo unívocamente con un 90% de probabilidad, y 4 si tomamos los menos populares. Por último, esta tesis se ha visto afectada por el estallido de la pandemia del COVID- 19, lo que ha contribuido al análisis de la evolución del mercado de la publicidad en línea con este periodo. La investigación demuestra que el mercado de la publicidad muestra una inelasticidad casi perfecta en la oferta y que cambió su composición debido a un cambio en el comportamiento en línea de los usuarios. También ilustra el potencial que tiene la utilización de los datos de los grandes servicios en línea, dado que ya tienen una alta tasa de adopción, y presenta un protocolo para la localización de contactos que han estado potencialmente expuestos a personas que direon positivo en COVID-19, en contraste con el fracaso de las nuevas aplicaciones de localización de contactos. En conclusión, la investigación de esta tesis muestra el impacto social y económico de la publicidad online y de los grandes servicios online en los usuarios. La metodología utilizada y desplegada sirve para poner de manifiesto y cuantificar los riesgos derivados de los datos personales en los servicios en línea. Presenta la necesidad de tales herramientas y metodologías en consonancia con la nueva legislación y los deseos de los usuarios. Siguiendo estas peticiones, en la búsqueda de transparencia y privacidad, esta tesis muestra soluciones y medidas fácilmente implementables para prevenir estos riesgos y capacitar al usuario para controlar su información personal. The web ecosystem is enormous, and overall it is sustained by an intangible attribute that mainly supports the majority of free services: the exploitation of personal information. Over the years, concerns on how services use personal data have increased and attracted the attention of media and users. This collection of personal information is the primary source of revenue on the Internet nowadays. Furthermore, on top of this, online advertising is the piece that supports it all. Without the existence of personal data in communion with online advertising, the Internet would probably not be the giant we know today. Online advertising is a very complex ecosystem in which multiple stakeholders take part. It is the motor that generates revenue on the web, and it has evolved in a few years to reach billions of users worldwide. While browsing, users generate valuable data about themselves that advertisers later use to offer them relevant products in which users could be interested. It is a two-way approach since advertisers pay intermediates to show ads to the public that is, in principle, most interested. However, this trading, sharing, and processing of personal data and behavior patterns, apart from opening up new advertising ways, expose users’ privacy. This incessant collection and commercialization of personal information usually fall behind an opaque wall, where the user often does not know what their data is used for. Privacy and transparency initiatives have increased over the years to empower the user in this business that moves billions of US dollars in revenue. Not surprisingly, after several scandals, such as the Facebook Cambridge Analytica scandal, businesses and regulators have joined forces to create transparency and protect users against the harmful practices derived from the use of their personal information. For instance, the General Data Protection Regulation (GDPR), is the most promising example of a data protection regulation, affecting all the member states of the European Union (EU), advocating for protecting users. The content of this thesis will use this legislation as a reference. For all these reasons, the purpose of this thesis is to provide tools and methodologies that reveal inappropriate uses of personal data by large companies in the online advertising ecosystem and create transparency among users, providing solutions to protect themselves. Thus, the content of this thesis offers design, analysis, and implementation of methodologies that measure online personal information’s social and economic impact on extensive Internet services. Mainly, it focuses on Facebook (FB), one of the largest social networks and services on the web, accounting with more than 2.8B Monthly Active Users (MAU) worldwide and generating only in online advertising revenue, more than $84B in 2020. First, this thesis presents a solution, in the form of a browser extension called Data Valuation Tool for Facebook users (FDVT), to provide users with a personalized, real-time estimation of the money they are generating for FB. By analyzing the number of ads and interactions in a session, the user gets information on their value within this social network. The add-on has had significant impact and adoption both by users, being installed more than 10k times since its public launch in October 2016, and media, appearing in more than 100 media outlets. Second, the study and research of the potential risks associated with processing users’ data should also follow the creation of these kinds of solutions. In this context, this thesis discovers and unveils striking results on the usage of personal information: (i) it quantifies the number of users affected by the usage of sensitive attributes used for advertising on FB, using as reference the definition of sensitive data from the GDPR. This thesis relies on the use of Natural Language Processing (NLP) to identify sensitive attributes, and it later uses the FB Ads Manager to retrieve the number of users assigned with this sensitive information. Two-thirds of FB users are affected by the use of sensitive personal data attributed to them. Moreover, the legislation seems not to affect this use of sensitive attributes from FB, and it presents severe risks to users. (ii) It models the number of non-Personal Identifiable Information (PII) attributes that are enough to uniquely identify an individual over a database of billions of users and proofs that reaching a single user is plausible even without knowing PII data of themselves. The results demonstrate that 22 interests at random from a user are enough to identify them uniquely with a 90% of probability, and 4 when taking the least popular ones. Finally, this thesis was affected by the outbreak of the COVID-19 pandemic what led to side contribute to the analysis of how the online advertising market evolved during this period. The research shows that the online advertising market shows an almost perfect inelasticity on supply and that it changed its composition due to a change in users’ online behavior. It also illustrates the potential of using data from large online services which already have a high adoption rate and presents a protocol for contact tracing individuals who have been potentially exposed to people who tested positive in COVID-19, in contrast to the failure of newly deployed contact tracing apps. In conclusion, the research for this thesis showcases the social and economic impact of online advertising and extensive online services on users. The methodology used and deployed is used to highlight and quantify the risks derived from personal data in online services. It presents the necessity of such tools and methodologies in line with new legislation and users’ desires. Following these requests, in the search for transparency and privacy, this thesis displays easy implementable solutions and measurements to prevent these risks and empower the user to control their personal information. This work was supported by the Ministerio de Educación, Cultura y Deporte, Spain, through the FPU Grant FPU16/05852, the Ministerio de Ciencia e Innovación, Spain, through the project ACHILLES Grant PID2019-104207RB-I00, the H2020 EU-Funded SMOOTH project under Grant 786741, and the H2020 EU-Funded PIMCITY project under Grant 871370. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: David Larrabeiti López.- Secretario: Gregorio Ignacio López López.- Vocal: Noel Crespi

Mención Internacional en el título de doctor Esta tesis pretende ayudar en la definición y el diseño de la quinta generación de redes de telecomunicaciones (5G) a través del modelado matemático de las diferentes cualidades que las caracterizan. En general, la ambición de estos modelos es realizar una optimización de las redes, ensalzando sus capacidades recientemente adquiridas para mejorar la eficiencia de los futuros despliegues tanto para los usuarios como para los operadores. El periodo de realización de esta tesis se corresponde con el periodo de investigación y definición de las redes 5G, y, por lo tanto, en paralelo y en el contexto de varios proyectos europeos del programa H2020. Por lo tanto, las diferentes partes del trabajo presentado en este documento cuadran y ofrecen una solución a diferentes retos que han ido apareciendo durante la definición del 5G y dentro del ámbito de estos proyectos, considerando los comentarios y problemas desde el punto de vista de todos los usuarios finales, operadores y proveedores. Así, el primer reto a considerar se centra en el núcleo de la red, en particular en cómo integrar tráfico fronthaul y backhaul en el mismo estrato de transporte. La solución propuesta es un marco de optimización para el enrutado y la colocación de recursos que ha sido desarrollado teniendo en cuenta restricciones de retardo, capacidad y caminos, maximizando el grado de despliegue de Unidades Distribuidas (DU) mientras se minimizan los agregados de las Unidades Centrales (CU) que las soportan. El marco y los algoritmos heurísticos desarrollados (para reducir la complexidad computacional) son validados y aplicados a redes tanto a pequeña como a gran (nivel de producción) escala. Esto los hace útiles para los operadores de redes tanto para la planificación de la red como para el ajuste dinámico de las operaciones de red en su infraestructura (virtualizada). Moviéndonos más cerca de los usuarios, el segundo reto considerado se centra en la colocación de servicios en entornos de nube y borde (cloud/edge). En particular, el problema considerado consiste en seleccionar la mejor localización para cada función de red virtual (VNF) que compone un servicio en entornos de robots en la nube, que implica restricciones estrictas en las cotas de retardo y fiabilidad. Los robots, vehículos y otros dispositivos finales proveen competencias significativas como impulsores, sensores y computación local que son esenciales para algunos servicios. Por contra, estos dispositivos están en continuo movimiento y pueden perder la conexión con la red o quedarse sin batería, cosa que reta aún más la entrega de servicios en este entorno dinámico. Así, el análisis realizado y la solución propuesta abordan las restricciones de movilidad y batería. Además, también se necesita tener en cuenta los aspectos temporales y los objetivos conflictivos de fiabilidad y baja latencia en el despliegue de servicios en una red volátil, donde los nodos de cómputo móviles actúan como una extensión de la infraestructura de cómputo de la nube y el borde. El problema se formula como un problema de optimización para colocación de VNFs minimizando el coste y también se propone un heurístico eficiente. Los algoritmos son evaluados de forma extensiva desde varios aspectos por simulación en escenarios que reflejan la realidad de forma detallada. Finalmente, el último reto analizado se centra en dar soporte a servicios basados en el borde, en particular, aprendizaje automático (ML) en escenarios del Internet de las Cosas (IoT) distribuidos. El enfoque tradicional al ML distribuido se centra en adaptar los algoritmos de aprendizaje a la red, por ejemplo, reduciendo las actualizaciones para frenar la sobrecarga. Las redes basadas en el borde inteligente, en cambio, hacen posible seguir un enfoque opuesto, es decir, definir la topología de red lógica alrededor de la tarea de aprendizaje a realizar, para así alcanzar el resultado de aprendizaje deseado. La solución propuesta incluye un modelo de sistema que captura dichos aspectos en el contexto de ML supervisado, teniendo en cuenta tanto nodos de aprendizaje (que realizan las computaciones) como nodos de información (que proveen datos). El problema se formula para seleccionar (i) qué nodos de aprendizaje e información deben cooperar para completar la tarea de aprendizaje, y (ii) el número de iteraciones a realizar, para minimizar el coste de aprendizaje mientras se garantizan los objetivos de error predictivo y tiempo de ejecución. La solución también incluye un algoritmo heurístico que es evaluado ensalzando una topología de red real y considerando tanto las tareas de clasificación como de regresión, y cuya solución se acerca mucho al óptimo, superando las soluciones alternativas encontradas en la literatura. This thesis aims to help in the definition and design of the 5th generation of telecommunications networks (5G) by modelling the different features that characterize them through several mathematical models. Overall, the aim of these models is to perform a wide optimization of the network elements, leveraging their newly-acquired capabilities in order to improve the efficiency of the future deployments both for the users and the operators. The timeline of this thesis corresponds to the timeline of the research and definition of 5G networks, and thus in parallel and in the context of several European H2020 programs. Hence, the different parts of the work presented in this document match and provide a solution to different challenges that have been appearing during the definition of 5G and within the scope of those projects, considering the feedback and problems from the point of view of all the end users, operators and providers. Thus, the first challenge to be considered focuses on the core network, in particular on how to integrate fronthaul and backhaul traffic over the same transport stratum. The solution proposed is an optimization framework for routing and resource placement that has been developed taking into account delay, capacity and path constraints, maximizing the degree of Distributed Unit (DU) deployment while minimizing the supporting Central Unit (CU) pools. The framework and the developed heuristics (to reduce the computational complexity) are validated and applied to both small and largescale (production-level) networks. They can be useful to network operators for both network planning as well as network operation adjusting their (virtualized) infrastructure dynamically. Moving closer to the user side, the second challenge considered focuses on the allocation of services in cloud/edge environments. In particular, the problem tackled consists of selecting the best the location of each Virtual Network Function (VNF) that compose a service in cloud robotics environments, that imply strict delay bounds and reliability constraints. Robots, vehicles and other end-devices provide significant capabilities such as actuators, sensors and local computation which are essential for some services. On the negative side, these devices are continuously on the move and might lose network connection or run out of battery, which further challenge service delivery in this dynamic environment. Thus, the performed analysis and proposed solution tackle the mobility and battery restrictions. We further need to account for the temporal aspects and conflicting goals of reliable, low latency service deployment over a volatile network, where mobile compute nodes act as an extension of the cloud and edge computing infrastructure. The problem is formulated as a cost-minimizing VNF placement optimization and an efficient heuristic is proposed. The algorithms are extensively evaluated from various aspects by simulation on detailed real-world scenarios. Finally, the last challenge analyzed focuses on supporting edge-based services, in particular, Machine Learning (ML) in distributed Internet of Things (IoT) scenarios. The traditional approach to distributed ML is to adapt learning algorithms to the network, e.g., reducing updates to curb overhead. Networks based on intelligent edge, instead, make it possible to follow the opposite approach, i.e., to define the logical network topology around the learning task to perform, so as to meet the desired learning performance. The proposed solution includes a system model that captures such aspects in the context of supervised ML, accounting for both learning nodes (that perform computations) and information nodes (that provide data). The problem is formulated to select (i) which learning and information nodes should cooperate to complete the learning task, and (ii) the number of iterations to perform, in order to minimize the learning cost while meeting the target prediction error and execution time. The solution also includes an heuristic algorithm that is evaluated leveraging a real-world network topology and considering both classification and regression tasks, and closely matches the optimum, outperforming state-of-the-art alternatives. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Pablo Serrano Yáñez-Mingot.- Secretario: Andrés García Saavedra.- Vocal: Luca Valcarenghi

Mención Internacional en el título de doctor While oceans occupy most of our planet, their exploration and conservation are one of the crucial research problems of modern time. Underwater localization stands among the key issues on the way to the proper inspection and monitoring of this significant part of our world. In this thesis, we investigate and tackle different challenges related to underwater ranging and localization. In particular, we focus on algorithms that consider underwater acoustic channel properties. This group of algorithms utilizes additional information about the environment and its impact on acoustic signal propagation, in order to improve the accuracy of location estimates, or to achieve a reduced complexity, or a reduced amount of resources (e.g., anchor nodes) compared to traditional algorithms. First, we tackle the problem of passive range estimation using the differences in the times of arrival of multipath replicas of a transmitted acoustic signal. This is a costand energy- effective algorithm that can be used for the localization of autonomous underwater vehicles (AUVs), and utilizes information about signal propagation. We study the accuracy of this method in the simplified case of constant sound speed profile (SSP) and compare it to a more realistic case with various non-constant SSP. We also propose an auxiliary quantity called effective sound speed. This quantity, when modeling acoustic propagation via ray models, takes into account the difference between rectilinear and non-rectilinear sound ray paths. According to our evaluation, this offers improved range estimation results with respect to standard algorithms that consider the actual value of the speed of sound. We then propose an algorithm suitable for the non-invasive tracking of AUVs or vocalizing marine animals, using only a single receiver. This algorithm evaluates the underwater acoustic channel impulse response differences induced by a diverse sea bottom profile, and proposes a computationally- and energy-efficient solution for passive localization. Finally, we propose another algorithm to solve the issue of 3D acoustic localization and tracking of marine fauna. To reach the expected degree of accuracy, more sensors are often required than are available in typical commercial off-the-shelf (COTS) phased arrays found, e.g., in ultra short baseline (USBL) systems. Direct combination of multiple COTS arrays may be constrained by array body elements, and lead to breaking the optimal array element spacing, or the desired array layout. Thus, the application of state-of-the-art direction of arrival (DoA) estimation algorithms may not be possible. We propose a solution for passive 3D localization and tracking using a wideband acoustic array of arbitrary shape, and validate the algorithm in multiple experiments, involving both active and passive targets. Part of the research in this thesis has been supported by the EU H2020 program under project SYMBIOSIS (G.A. no. 773753). This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Paul Daniel Mitchell.- Secretario: Antonio Fernández Anta.- Vocal: Santiago Zazo Bello

Mención Internacional en el título de doctor With the arrival of new generation mobile networks, we currently observe a paradigm shift, where monolithic network functions running on dedicated hardware are now implemented as software pieces that can be virtualized on general purpose hardware platforms. This paradigm shift stands on the softwarization of network functions and the adoption of virtualization techniques. Network Function Virtualization (NFV) comprises softwarization of network elements and virtualization of these components. It brings multiple advantages: (i) Flexibility, allowing an easy management of the virtual network functions (VNFs) (deploy, start, stop or update); (ii) efficiency, resources can be adequately consumed due to the increased flexibility of the network infrastructure; and (iii) reduced costs, due to the ability of sharing hardware resources. To this end, multiple challenges must be addressed to effectively leverage of all these benefits. Network Function Virtualization envisioned the concept of virtual network, resulting in a key enabler of 5G networks flexibility, Network Slicing. This new paradigm represents a new way to operate mobile networks where the underlying infrastructure is "sliced" into logically separated networks that can be customized to the specific needs of the tenant. This approach also enables the ability of instantiate VNFs at different locations of the infrastructure, choosing their optimal placement based on parameters such as the requirements of the service traversing the slice or the available resources. This decision process is called orchestration and involves all the VNFs withing the same network slice. The orchestrator is the entity in charge of managing network slices. Hands-on experiments on network slicing are essential to understand its benefits and limits, and to validate the design and deployment choices. While some network slicing prototypes have been built for Radio Access Networks (RANs), leveraging on the wide availability of radio hardware and open-source software, there is no currently open-source suite for end-to-end network slicing available to the research community. Similarly, orchestration mechanisms must be evaluated as well to properly validate theoretical solutions addressing diverse aspects such as resource assignment or service composition. This thesis contributes on the study of the mobile networks evolution regarding its softwarization and cloudification. We identify software patterns for network function virtualization, including the definition of a novel mobile architecture that squeezes the virtualization architecture by splitting functionality in atomic functions. Then, we effectively design, implement and evaluate of an open-source network slicing implementation. Our results show a per-slice customization without paying the price in terms of performance, also providing a slicing implementation to the research community. Moreover, we propose a framework to flexibly re-orchestrate a virtualized network, allowing on-the-fly re-orchestration without disrupting ongoing services. This framework can greatly improve performance under changing conditions. We evaluate the resulting performance in a realistic network slicing setup, showing the feasibility and advantages of flexible re-orchestration. Lastly and following the required re-design of network functions envisioned during the study of the evolution of mobile networks, we present a novel pipeline architecture specifically engineered for 4G/5G Physical Layers virtualized over clouds. The proposed design follows two objectives, resiliency upon unpredictable computing and parallelization to increase efficiency in multi-core clouds. To this end, we employ techniques such as tight deadline control, jitter-absorbing buffers, predictive Hybrid Automatic Repeat Request, and congestion control. Our experimental results show that our cloud-native approach attains > 95% of the theoretical spectrum efficiency in hostile environments where stateof- the-art architectures collapse. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Francisco Valera Pintor.- Secretario: Vincenzo Sciancalepore.- Vocal: Xenofon Foukas

Technological paradigms such as Software Defined Networking, Network Function Virtualization and Network Slicing are altogether offering new ways of providing services. This process is widely known as Network Softwarization, where traditional operational networks adopt capabilities and mechanisms inherit form the computing world, such as programmability, virtualization and multi-tenancy. This adoption brings a number of challenges, both from the technological and operational perspectives. On the other hand, they provide an unprecedented flexibility opening opportunities to developing new services and new ways of exploiting and consuming telecom networks. This Thesis first overviews the implications of the progressive introduction of network softwarization in operational networks for later on detail some advances at different levels, namely architectural, service and transport levels. It is done through specific exemplary use cases and evolution scenarios, with the goal of illustrating both new possibilities and existing gaps for the ongoing transition towards an advanced future mode of operation. This is performed from the perspective of a telecom operator, paying special attention on how to integrate all these paradigms into operational networks for assisting on their evolution targeting new, more sophisticated service demands. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Eduardo Juan Jacob Taquet.- Secretario: Francisco Valera Pintor.- Vocal: Jorge López Vizcaíno

El ecosistema de servicios digitales disponibles para los dispositivos domésticos ha estado creciendo durante los últimos años. Ese crecimiento no solo es causado por el aumento significativo de los contenidos digitales disponibles, sino también por el aumento de los dispositivos conectados a Internet que son capaces de consumir estos contenidos. La mayoría de estos dispositivos suelen ser dispositivos móviles y otros dispositivos personales de consumo los cuales se están convirtiendo en los clientes preferidos. Actualmente, la mayoría de los servicios dirigidos a dispositivos heterogéneos son servicios Over-the-top (OTT). Estos servicios OTT permiten que varios dispositivos consuman el mismo contenido, ya que no hay necesidad de una infraestructura de servicio subyacente. Sin embargo, estos servicios funcionarían mejor utilizando una infraestructura dedicada que proporciona una mayor experiencia de usuario mientras que requiere menos recursos. Tomando en cuenta este panorama se puede pensar en un futuro escenario caracterizado por una gran diversidad de dispositivos, de usuarios potenciales y un gran número de servicios disponibles por lo que sería conveniente disponer de infraestructuras fiables y robustas y además poder dar soporte a cualquier tipo de servicio. Dado que la infraestructura física puede ser costosa de implementar, lo más lógico sería considerar que este ecosistema de servicios digitales va a requerir el uso de infraestructuras dinámicas y escalables. Afortunadamente, el paradigma del Cloud Computing, y sus alternativas como el Edge Computing y el Fog Computing, permiten crear instancias dinámicas de servicios y elementos de red haciendo frente a los requisitos antes mencionados mientras reduce los costes de inversión y operación. Sin embargo, el paradigma de cloud computing tiene varios problemas que requieren especial atención como: Data Lock-In (o bloqueo de proveedor) y la interoperabilidad entre proveedores de infraestructura. Hoy en día, los orquestadores de computación en la nube son bastante heterogéneos y carecen de interfaces estándar que permitan realizar despliegues automatizados. Por lo tanto, una definición de infraestructura diseñada para un proveedor de infraestructura en la nube determinado es más que probable que no se pueda crear una instancia en otros. Por lo tanto, se requieren mecanismos de configuración flexibles, administración avanzada, procesamiento descentralizado y auto-organización. En esta tesis, los autores proponen una arquitectura que amplía las funciones de red definidas en las especificaciones y estándares de NFV para incorporar funcionalidades de mayor nivel. El objetivo es ampliar la virtualización de las funciones de red con la creación de instancias de servicios en dispositivos de consumo. Los autores proponen ampliar la idea de NFV en lugar de proponer un nuevo diseño porque NFV tiene una gran aceptación en la industria y por eso contribuiremos ampliando la arquitectura NFV. De esta manera, los proveedores de servicios tendrán un control adicional del software del sistema al mover algunos servicios, tradicionalmente ubicados en los dispositivos, a la nube. Uno de los mayores problemas encontrados para extender la idea al mayor número de infraestructuras se debe a las particularidades de cada una de ellas que las hacen diferentes. Para superar esto, proponemos un modelo estructurado de información para declarar componentes, conexiones y configuraciones, que permite a la solución de arquitectura propuesta en la tesis crear instancias de un determinado conjunto de servicios en diferentes proveedores de infraestructuras en la nube. Para que esta configuración pueda ser aplicada a varios niveles en las infraestructuras, este trabajo propone utilizar un modelo segmentado por capas de software. Por lo tanto, los orquestadores pueden implementar e interconectar partes de software de una manera sencilla, independientemente del proveedor del entorno. En esta tesis se propone un modelo arquitectónico que permite construir un ecosistema virtual en entornos domésticos que requieren muchas y complejas tareas de configuración. El uso de mecanismos automáticos de gestión de software garantiza la correcta configuración y despliegue de servicios en diferentes plataformas acelerando el proceso y reduciendo la probabilidad de cometer errores de configuración. A continuación se presentan las principales contribuciones en este contexto: Diseño de un ecosistema digital doméstico e integrado. Modelo de arquitectura flexible y adaptable. Automatización en la configuración de infraestructuras y despliegue de servicios. Optimización de dependencias y del despliegue de servicios software. Modelado de la información del software de infraestructura. Orquestación del software basado en herramientas de la IaC. En resumen, el trabajo presentado en esta tesis contribuye a avanzar en el desarrollo de los ecosistemas domésticos digitales con el objetivo de proporcionar a los proveedores de servicios nuevas herramientas que les haga más fácil la gestión de los sistemas a un precio más bajo. Los consumidores de los servicios podrán gestionar de manera sencilla ecosistemas digitales donde la frontera entre lo virtual, los servicios contratados y sus dispositivos domésticos y móviles forman un todo integrado y perfectamente orquestado. Parte del contenido se desarrolló como línea de investigación en varios proyectos de I+D, denominados: - RAUDOS2 Red Interactiva Multiplataforma de Distribución de Contenidos Audiovisuales (TSI-020302-2010-67) (Proy. Nacional) - HAUS-Hogar digital y contenidos Audiovisuales adaptados a los USuarios (IPT-2011-1049-430000) (Proy. Nacional) - IRENE - Incentivación del reciclaje de envases con NFC en España (PT-2012-1036-370000) (Proy. Nacional) - COMINN - Centro comercial interactivo con interacción natural (IPT- 2012-0883-430000) (Proy. Nacional) - REMEDISS - Red médica social sensorizada (IPT-2012-0882-430000) - OSAMI-Commons Open Source Infrastructure Ambient Inteligence Commons (TSI-020400-2009-92) (Proy. Internacional) Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Fernando Pescador del Oso.- Secretario: Andrés Marín López.- Vocal: Francisco José Bellido Outeiriño

Mención Internacional en el título de doctor In recent years, we have experienced several social media blackouts, which have shown how much our daily experiences depend on high-quality communication services. Blackouts have occurred because of technical problems, natural disasters, hacker attacks or even due to deliberate censorship actions undertaken by governments. In all cases, the spontaneous reaction of people consisted in finding alternative channels and media so as to reach out to their contacts and partake their experiences. Thus, it has clearly emerged that infrastructured networks—and cellular networks in particular—are well engineered and have been extremely successful so far, although other paradigms should be explored to connect people. The most promising of today’s alternative paradigms is Device-to-Device (D2D) because it allows for building networks almost freely, and because 5G standards are (for the first time) seriously addressing the possibility of using D2D communications. In this dissertation I look at opportunistic D2D networking, possibly operating in an infrastructure-less environment, and I investigate several schemes through modeling and simulation, deriving metrics that characterize their performance. In particular, I consider variations of the Floating Content (FC) paradigm, that was previously proposed in the technical literature. Using FC, it is possible to probabilistically store information over a given restricted local area of interest, by opportunistically spreading it to mobile users while in the area. In more detail, a piece of information which is injected in the area by delivering it to one or more of the mobile users, is opportunistically exchanged among mobile users whenever they come in proximity of one another, progressively reaching most (ideally all) users in the area and thus making the information dwell in the area of interest, like in a sort of distributed storage. While previous works on FC almost exclusively concentrated on the communication component, in this dissertation I look at the storage and computing components of FC, as well as its capability of transferring information from one area of interest to another. I first present background work, including a brief review of my Master Thesis activity, devoted to the design, implementation and validation of a smartphone opportunistic information sharing application. The goal of the app was to collect experimental data that permitted a detailed analysis of the occurring events, and a careful assessment of the performance of opportunistic information sharing services. Through experiments, I showed that many key assumptions commonly adopted in analytical and simulation works do not hold with current technologies. I also showed that the high density of devices and the enforcement of long transmission ranges for links at the edge might counter-intuitively impair performance. The insight obtained during my Master Thesis work was extremely useful to devise smart operating procedures for the opportunistic D2D communications considered in this dissertation. In the core of this dissertation, initially I propose and study a set of schemes to explore and combine different information dissemination paradigms along with real users mobility and predictions focused on the smart diffusion of content over disjoint areas of interest. To analyze the viability of such schemes, I have implemented a Python simulator to evaluate the average availability and lifetime of a piece of information, as well as storage usage and network utilization metrics. Comparing the performance of these predictive schemes with state-of-the-art approaches, results demonstrate the need for smart usage of communication opportunities and storage. The proposed algorithms allow for an important reduction in network activity by decreasing the number of data exchanges by up to 92%, requiring the use of up to 50% less of on-device storage, while guaranteeing the dissemination of information with performance similar to legacy epidemic dissemination protocols. In a second step, I have worked on the analysis of the storage capacity of probabilistic distributed storage systems, developing a simple yet powerful information theoretical analysis based on a mean field model of opportunistic information exchange. I have also extended the previous simulator to compare the numerical results generated by the analytical model to the predictions of realistic simulations under different setups, showing in this way the accuracy of the analytical approach, and characterizing the properties of the system storage capacity. I conclude from analysis and simulated results that when the density of contents seeded in a floating system is larger than the maximum amount which can be sustained by the system in steady state, the mean content availability decreases, and the stored information saturates due to the effects of resource contention. With the presence of static nodes, in a system with infinite host memory and at the mean field limit, there is no upper bound to the amount of injected contents which a floating system can sustain. However, as with no static nodes, by increasing the injected information, the amount of stored information eventually reaches a saturation value which corresponds to the injected information at which the mean amount of time spent exchanging content during a contact is equal to the mean duration of a contact. As a final step of my dissertation, I have also explored by simulation the computing and learning capabilities of an infrastructure-less opportunistic communication, storage and computing system, considering an environment that hosts a distributed Machine Learning (ML) paradigm that uses observations collected in the area over which the FC system operates to infer properties of the area. Results show that the ML system can operate in two regimes, depending on the load of the FC scheme. At low FC load, the ML system in each node operates on observations collected by all users and opportunistically shared among nodes. At high FC load, especially when the data to be opportunistically exchanged becomes too large to be transmitted during the average contact time between nodes, the ML system can only exploit the observations endogenous to each user, which are much less numerous. As a result, I conclude that such setups are adequate to support general instances of distributed ML algorithms with continuous learning, only under the condition of low to medium loads of the FC system. While the load of the FC system induces a sort of phase transition on the ML system performance, the effect of computing load is more progressive. When the computing capacity is not sufficient to train all observations, some will be skipped, and performance progressively declines. In summary, with respect to traditional studies of the FC opportunistic information diffusion paradigm, which only look at the communication component over one area of interest, I have considered three types of extensions by looking at the performance of FC: over several disjoint areas of interest; in terms of information storage capacity; in terms of computing capacity that supports distributed learning. The three topics are treated respectively in Chapters 3 to 5. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Claudio Ettori Casetti.- Secretario: Antonio de la Oliva Delgado.- Vocal: Christoph Sommer

Mención Internacional en el título de doctor Este trabajo se ha realizado bajo la ayuda concedida por la Comunidad de Madrid en la Convocatoria de 2017 de Ayudas para la Realización de Doctorados Industriales en la Comunidad de Madrid (Orden 3109/2017, de 29 de agosto), con referencia IND2017/TIC-7732. This work was partly funded by the European Commission under the European Union’s Horizon 2020 program - grant agreement number 815074 (5G EVE project). The Ph.D thesis solely reflects the views of the author. The Commission is not responsible for the contents of this Ph.D thesis or any use made thereof. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Antonio de la Oliva Delgado.- Secretaria: Elisa Rojas Sánchez.- Vocal: David Manuel Gutiérrez Estévez

Millimeter-wave (mmWave) is one of the key technologies that enables the next wireless generation. mmWave offers a much higher bandwidth than sub-6GHz communications which allows multi-gigabit-per-second rates. This also alleviates the scarcity of spectrum at lower frequencies, where most devices connect through sub-6GHz bands. However new techniques are necessary to overcome the challenges associated with such high frequencies. Most of these challenges come from the high spatial attenuation at the mmWave band, which requires new paradigms that differ from sub-6GHz communications. Most notably mmWave telecommunications are characterized by the need to be directional in order to extend the operational range. This is achieved by using electronically steerable antenna arrays, that focus the energy towards the desired direction by combining each antenna element constructively or destructively. Additionally, most of the energy comes from the Line Of Sight (LOS) component which gives mmWave a quasi-optical behaviour where signals can reflect off walls and still be used for communication. Some other challenges that directional communications bring are mobility tracking, blockages and misalignments due to device rotation. The IEEE 802.11ad amendment introduced wireless telecommunications in the unlicensed 60 GHz band. It is the first standard to address the limitations of mmWave. It does so by introducing new mechanisms at the Medium Access Control (MAC) and Physical (PHY) layers. It introduces multi-band operation, relay operation mode, hybrid channel access scheme, beam tracking and beam forming among others. In this thesis we present a series of works that aim to improve mmWave telecommunications. First we give an overview of the intrinsic challenges of mmWave telecommunications, by explaining the modifications to the MAC and PHY layers. This sets the base for the rest of the thesis. Then do a comprehensive study on how mmWave behaves with existing technologies, namely TCP. TCP is unable to distinguish losses caused by congestion or by transmission errors caused by channel degradation. Since mmWave is affected by blockages more than sub-6GHz technologies, we propose a set of parameters that improve the channel quality even for mobile scenarios. The next job focuses on reducing the initial access overhead of mmWave by using sub-6GHz information to steer towards the desired direction. We start this work by doing a comprehensive High Frequency (HF) and Low Frequency (LF) correlation, analyzing the similarity of the existing paths between the two selected frequencies. Then we propose a beam steering algorithm that reduces the overhead to one third of the original time. Once we have studied how to reduce the initial access overhead, we propose a mechanism to reduce the beam tracking overhead. For this we propose an open platform based on a Field Programmable Gate Arrays (FPGA) where we implement an algorithm that completely removes the need to train on the Station (STA) side. This is achieved by changing beam patterns on the STA side while the Access Point (AP) is sending the preamble. We can change up to 10 beam patterns without losing connection and we reduce the overhead by a factor of 8.8 with respect to the IEEE 802.11ad standard. Finally we present a dual band location system based on Commercial-Off-The-Shelve (COTS) devices. Locating the STA can improve the quality of the channel significantly, since the AP can predict and react to possible blockages. First we reverse engineer existing 60 GHz enabled COTS devices to extract Channel State Information (CSI) and Fine Timing Measurements (FTM) measurements, from which we can estimate angle and distance. Then we develop an algorithm that is able to choose between HF and LF in order to improve the overall accuracy of the system. We achieve less than 17 cm of median error in indoor environments, even when some areas are Non Line Of Sight (NLOS). This work has been supported by IMDEA Networks Institute. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Matthias Hollick.- Secretario: Vincenzo Mancuso.- Vocal: Paolo Casari

Mención Internacional en el título de doctor Conventional communications between human mobile users are not the only type of activities envisioned for post-5G and future 6G networks. Traditional Human-to- Human (H2H) communications are now joined by Human-to-Machine (H2M), and Machine-to-Machine (M2M) communications, in a Tactile Internet world. 6G is expected to put an end to smartphone-centric era, introducing new system paradigms. New services and applications such as telepresence, remote control of haptic machines, online schools/working, remote surgery, Multi-access Edge Computing (MEC) are envisioned to be part of the future daily life. Cloud-based services (including the cloud-based signal processing of radio waveforms), video traffic, enterprise applications, and virtualization have led to a rise of the traffic demand in metropolitan areas. Traditional cellular architectures are becoming exhausted and unable to cope with the requested performance and new approaches such as the Cloud Radio Access Networks (C-RAN) appear as natural replacements. This technology suggests the separation between the radio and the processing equipment that now will be located at different places of the network. However, all the implementation details about how to meet the bandwidth, latency, jitter or even how to measure these parameters remain open. In this thesis, we firstly assess real world traditional deployments for C-RAN like ring-star topologies implemented with Ethernet switching equipment. We derive the theoretical expressions for the propagation and queueing delays, assuming G/G/1 queueing models. Furthermore, the properties of the fronthaul streams are examined, particularly regarding the aggregation of functional Split B flows in the same fronthaul network. Additionally, the G/G/1 queueing model is further extended with the Kingman’s Exponential Law of Congestion, paying special attention to the percentile values of the queueing delay. High queueing delay percentiles are proposed as the key metric for the fronthaul network dimensioning. Namely, the aggregation of eCPRI flows (Splits IU and IID) in a packet aggregator is explored for different values of the radio channel’s bandwidth. The use of extreme latency percentiles is later explored as a useful tool for the design of fronthaul networks. The G/G/1 and N*D/D/1 queueing models are tested as modeling tools and compared with simulation to assess the tightness of their predictions in the extreme latency percentiles scenarios. The results support that the N*D/D/1 queue is able to model the behavior of a packet-switch fronthaul aggregator using the eCPRI standard for 5G New Radio (NR) fronthaul streams in a more accurate way and can be used as a tool to dimension the length of the links. This better modeling of the fronthaul translates into a finer implementation of C-RAN. By interpreting the gap betweenthis estimation and maximum worst-case delay as an extra delay budget, the fronthaul links’ lengths can be increased by 60% and 10% for 50 MHz and 100 MHz NR channels, respectively, while keeping the latency budget and frame loss ratio within the IEEE 802.1CM limits. Recent research efforts focus on the development of new photonic technologies, enabling capacities of Tb/s for the metropolitan area network. Sliceable Bandwidth Variable Transceivers (S-BVTs) appear as a promising technology to achieve this goal. The newadvancements in S-BVTs suggest that it is a cost-effective, flexible, and highperformance technology to improve metropolitan networks. In this thesis, we show their advantages over fixed transceivers in edge computing and content delivery network scenarios. Particularly, this work studies the dimensioning of edge data center resources from the perspective of achieving the blocking probabilities as low as 10[elevado a -6]. Two different scenarios are analyzed: the overflow over a paired samelevel data centers and the overflow over a centralized site. The results show that a proper distribution of computing resources in the centralized overflow approach can outperform the costs of a distributed strategy, requiring fewer processors and much smaller data centers at the central office. Moreover, different strategies to implement cost-effective CDN caching are presented. Fixed and bandwidth variable solutions are proposed and compared against each other in three different topologies, from the perspective of availability. Those solutions that make us of S-BVTs show many advantages in terms of scalability, low wavelength occupancy, and the reduced number of IT resources needed to support the same availability. Finally, this thesis addresses the characterization of the fronthaul traffic so that we may support its coexistence with other types of traffic and services. A Heterogeneous Cloud Radio Access Network (H-CRAN) serves as the reference network comprising all the envisioned traffic types, services, and data rates. A network simulator and its building blocks are described and presented as the tool to assess the performance of the proposed solution. This description includes the basics and the building blocks of the simulator (channel and interference models) as well as the definition of many concepts such as the heterogeneous networks, the construction of the coverage maps, the uplink and downlink decoupling, etc. A global optimization problem is proposed for the entire network, supporting three access technologies: LTE-A, WiFi, and C-RAN. A distributed and scalable operation method is proposed to orchestrate all the above components and to enable a seamless and flexible operation. The distributed algorithm is computationally simple as it only requires broadcasting the Lagrange multipliers that arise while solving the optimization problem. These, work as a sort of price indicator, transmitting information about the state and needs of base stations and users at any particular time. This solution is able to organize the user association and network resources in the appropriate way so as to comply with the delay requirements of each traffic type. As a desirable side effect, the energy consumption of the mobile nodes is also optimized. This work was supported in part by the Spanish Ministry of Education and Vocational Training under the FPU Grant number FPU16/01760, in part by the Spanish National TEXEO Project under Grant TEC2016-80339-R, in part by the H2020 EU-Funded BlueSpace Project under Grant 762055. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Luca Valcarenghi.- Secretario: Ángel Cuevas Rumín.- Vocal: Ramón Durán Barroso

Mención Internacional en el título de doctor Several business are nowadays becoming more and more aware of the potential that lies beneath Big Data. From social media ‘titans’ and healthcare companies, to the mobile industry that propelled them, 5th Generation mobile network (5G) will be a fundamental factor that will drive our new reality. Current mobile service usage has been explored for data-driven organizational growth in the touristic sector, as it allows forecasting hotel occupancy rates or targeting customersHowever, mobile data is continuously increasing, and therefore it is enormously important to analyze such data for networking purposes. Previous Ericsson Mobility Reports claimed that by the end of 2022 the total monthly traffic associated with mobile devices would be 77 exabytes (EB), representing 20% of the total Internet Protocol (IP) traffic around the world. They also declare that 50 EB/month would come from 2nd Generation mobile network (2G), 3rd Generation mobile network (3G), and 4th Generation mobile network (4G) devices. In terms of 5G subscriptions, it is expected to reach up to 2.8 billion subscriptions globally by the end of 2025, accounting for about 30% of total mobile subscriptions. Experts stake out that by the year 2020, 1.7 megabytes of data will be generated every second for every person on the planet, forcing the network to evolve and adapt to challenging new demands. Network providers not only deal with the deployment of the required resources to support this growth, but also the potential newcomers into the business, and the stringent conditions of the distinct services to be provided. One of the features proposed to face this dilemma is using the network slicing technique. It allows to transform and orchestrate a 5G network by creating multiple logical instances (i.e., slices) on top of it, while Big Data would provide the specifications of the services’ traffic dynamics to be served. In this way, operators achieve the best allocation of resources. This thesis contributes to the ongoing Network Slicing research, assessing a nationwide scenario. Our results show mobile traffic similarities and differences across time, space, and frequency domains, whereas we intend for distinct service clusterizations that would enhance the network efficiency in terms of resource management. For instance, we show that benefits are achieved when considering the top 10 consuming Network Slice (NS). In addition, we could observe mobile service similarities in the spatial domain, while the spectral and time domains open the door for wavelets uncertainty, where we point future directions to address this research branch. Moreover, we propose two data-driven algorithms that shed light on the trade-off between complexity and multiplexing efficiency derived from the network slice specifications, both exhibiting promising performances (e.g., leading to a new architecture for traffic balancing in the cloud and edge clusters, with 60% and 400% gain in efficiency respectively and 1/3 of dedicated resources). Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Antonio de la Oliva Delgado.- Secretario: Andrés García Saavedra.- Vocal: Andra Elena Lutu

Mención Internacional en el título de doctor La publicidad online ha evolucionado hasta convertirse en un componente clave del Internet que conocemos hoy en día. Es un ecosistema muy complejo, que logra llegar a billones de usuarios en un corto período de tiempo. Tiene cobertura global, y es capaz de llegar a audiencias específicas basadas en aspectos demográficos, geográficos y de comportamiento. Las capacidades que ofrece el ecosistema de la publicidad online han abierto una nueva era en la investigación que ha atraído el interés de la comunidad científica. Esta tesis, aprovecha la naturaleza de la publicidad online y construye una novedosa metodología capaz de insertar código JavaScript en un anuncio, que se ejecuta cada vez que se muestra en el dispositivo de un usuario. Esta metodología abre nuevas oportunidades para realizar medidas. En concreto, esta metodología se aplica para dos propósitos diferentes en esta tesis: (1) Realizar medidas de red desde la perspectiva del usuario final, y (2) Auditar la transparencia del ecosistema de la publicidad online desde la perspectiva de los anunciantes. En el contexto de las medidas de Internet, esta metodología se implementa en una soluciónllamada AdTag. Se discute y evalúa su diseño -incluyendo factores técnicos, de despliegue y económicos- y su potencial para analizar una amplia gama de aspectos de la conectividad a Internet desde el navegador. Se realizan varios experimentos que prueban la capacidad de AdTag para llegar a millones de nodos en un corto período de tiempo. Además, también se demuestra la posibilidad de seleccionar los nodos de medidas en función de su ubicación geográfica. En esta tesis, mostramos la utilidad de AdTag para realizar medidas de red en dos casos de uso específicos. Primero, estudiamos la infraestructura DNS, uno de los sistemas más críticos de Internet. Nuestro análisis aborda cuestiones como comprender la verdadera infraestructura DNS configurada por los ISP, y entender las opciones DNS de los usuarios finales, ya sea que utilicen los resolvers de los ISP privados o establezcan DNS resolvers de terceros, para mejorar la seguridad y el rendimiento de la web. Aprovechando la escala que ofrece el ecosistema de la publicidad online, se han lanzado dos campañas de publicidad que han conseguido más de 3 millones de resoluciones DNS, que permiten la identificación y el estudio de más de 76k DNS resolvers cubriendo más de 25k ASes en 178 países. El análisis de los datos proporciona nuevos conocimientos sobre la infraestructura DNS, como las preferencias de los usuarios con respecto a terceros. Nuestros resultados indican que el 13% de los usuarios utilizan proveedores DNS de terceros (como Google, OpenDNS, Level 3, y Cloudflare). Además, esta investigación detecta diferentes decisiones de despliegue de muchos ISP, que proporcionan acceso a redes tanto móviles como fijas, que separan la infraestructura DNS que sirve a cada tipo de red de acceso. El segundo caso de uso considerado consiste en analizar el escenario del mercado de navegadores mediante medidas activas. Aprovechamos AdTag para desarrollar una plataforma de medidas activa para obtener la marca y la versión del dispositivo que recibe el anuncio. Demostramos que la muestra obtenida con nuestra metodología es muy similar a la que ofrecen las técnicas de vanguardia basadas en medidas pasivas. Sin embargo, nuestra solución presenta algunas ventajas con respecto a las soluciones pasivas: la capacidad de llevar a cabo medidas dirigidas geográfica y demográficamente, además de su accesibilidad a un grupo más amplio de científicos y profesionales. El rendimiento, la precisión y las capacidades de esta metodología se analizan a través de experimentos reales que, en total, produjeron más de 6M de medidas. La falta de transparencia en el ecosistema de la publicidad online motiva la segunda parte de esta tesis. En particular, hemos desarrollado Q-Tag, una novedosa metodología que sirve para auditar las métricas de calidad de la publicidad online para que los anunciantes puedan obtener información fiable sobre el desempeño real de sus campañas publicitarias. La primera versión de Q-Tag fue desplegada en Google AdWords. Los resultados revelan que AdWords parece proporcionar información incompleta a los anunciantes. En particular, muestran que: (i) AdWords no informó sobre el 57% de los publishers, en los que se mostraron impresiones de anuncios de nuestras campañas, (ii) AdWords informa sobre una gran fracción de impresiones contextualmente significativas, basadas en criterios (no revelados) distintos del tema de los publishers, (iii) una mayor inversión en CPM no conduce a que las impresiones se entreguen a publishers más populares, (iv) AdWords no ofrece un control predeterminado sobre el frequency cap (límite de impresiones por usuario), (v) alrededor del 10% de las impresiones de anuncios en dos de las campañas se entregaron a IPs de Data Centers. La segunda versión de Q-Tag fue desarrollada para medir la métrica de viewability. Esta métrica estándar sirve para evaluar si una impresión de un anuncio ha sido vista o no por un usuario. Q-Tag ha sido desplegado en producción por un Demand Side Platform (DSP) (Plataforma del lado de la demanda) para medir el índice de visibilidad de las campañas publicitarias. Aprovechando la infraestructura de este DSP, se ha comparado el rendimiento de Q-Tag con una solución comercial. Ambas técnicas informan de una viewability global similar del 50% (es decir, el 50% de las impresiones cumplen con el estándar de viewability y por lo tanto se consideran vistas). Sin embargo, Q-Tag es capaz de medir la métrica de viewability en el 93% de los anuncios servidos por el DSP a diferencia del 74% de los anuncios medidos por la solución comercial. En resumen, la investigación realizada en esta tesis muestra el potencial de la metodología de medidas a gran escala basada en anuncios, que ofrece un mayor rango de posibilidades más allá de las presentadas en esta tesis. Una metodología que puede desentrañar diferentes aspectos de la infraestructura y el rendimiento de Internet desde la perspectiva del usuario final, así como proporcionar una herramienta independiente para que los anunciantes midan la calidad de sus campañas publicitarias. Online advertising has evolved into a key component of the Internet we know today. It is a very complex ecosystem that accomplishes to reach billions of users in a short period of time. It has global coverage, and it is able to target specific audiences based on demographic, geographic, and behavioral aspects. The capabilities offered by the online advertising ecosystem have opened a new era in research that has attracted the interest of the scientific community. This thesis leverages the nature of online advertising and builds a novel methodology capable of inserting JavaScript code into an ad that runs every time it is displayed on a user’s device. This methodology opens up new measurement opportunities. Specifically, this methodology is applied for two different purposes in this thesis: (1) Performing network measurements from the end-user perspective, and (2) Auditing the transparency of the online advertising ecosystem from the advertisers’ perspective. In the context of Internet measurements, this methodology is implemented in a solution referred to as AdTag. Its design - including technical, deployability, and economic factors – and its potential to analyze a wide range of aspects of Internet connectivity from the browser are discussed and evaluated. Several experiments are performed that prove the ability of AdTag to reach millions of nodes in a short period of time. Furthermore, the possibility of selecting the measurement nodes based on its geographical location is also demonstrated. In this thesis, we showcase the utility of AdTag to conduct network measurements in two specific use cases. First, we study the DNS infrastructure, one of the most critical Internet systems. Our analysis addresses issues such us grasping the real DNS infrastructure configured by the ISPs, and understanding the end-users DNS choices, whether they use private ISPs’ resolvers or establish third-party DNS resolvers, to improve security and web performance. Harnessing the scale offered by the online advertising ecosystem, two ad campaigns have been launched, triggering more than 3M DNS lookups, which allow the identification and study of more than 76k recursive DNS resolvers supporting more than 25k eyeball ASes in 178 countries. The data analysis provides new insights into the DNS infrastructure, such as user preferences towards third-parties. Our results indicate that 13% of users use third-party DNS providers (such as Google, OpenDNS, Level 3, and Cloudflare). Besides, this research detects different deployment decisions of many ISPs that provide both mobile and fixed access networks to separate the DNS infrastructure that serves each access technology type. The second considered use case consists of analyzing the browser market landscape with active measurements. We leverage AdTag to develop an active measurement platform to obtain the brand and the version of the device receiving the ad. We prove that the landscape picture obtained with our methodology is very similar to that offered by state-of-the-art techniques based on passive measurements. However, our solution presents some advantages over passive solutions: the ability to conduct geographically and demographically targeted measurements and its accessibility to a larger group of scientists and practitioners. The performance, accuracy, and capabilities of this methodology are analyzed through real experiments that, in total, produced more than 6M measurements. The lack of transparency in the online advertising ecosystem motivates the second part of this thesis. In particular, we have developed Q-Tag, a novel methodology that serves to audit reported quality metrics so that advertisers can obtain trustable information about the real performance of their advertising campaigns. The first version of Q-Tag was deployed in Google AdWords. The results reveal that AdWords seems to provide incomplete information to advertisers. In particular, they show that: (i) AdWords did not report 57% of the publishers where ad impressions from our campaigns were delivered, (ii) AdWords reports a large fraction of contextually significant impressions based on (undisclosed) criteria other than publisher’s theme, (iii) higher CPM investment does not lead to impressions being delivered to more popular publishers, (iv) AdWords does not offer default control of frequency cap (limit of impressions per user), (v) about 10% of ad impressions in two of the campaigns were delivered to IPs from Data Centers. The second version of Q-Tag was developed to measure the viewability metric. This standard metric serves to assess whether an ad impression was viewed or not by a user. Q-Tag has been deployed in production by a Demand Side Platform (DSP) to measure the viewability rate of the ad campaigns. Taking advantage of the infrastructure of this DSP, the performance of Q-Tag has been compared with a commercial solution. Both techniques report a similar overall viewability rate of 50% (i.e.,, 50% of the ad impressions meet the viewability standard and thus are considered viewed). However, Q-Tag is able to measure the viewability metric in 93% of the ads served by the DSP, unlike 74% of the ads measured by the commercial solution. In summary, the research conducted in this thesis showcases the potential of the proposed large-scale ad-based measurement. It offers a wider range of possibilities beyond those presented in this thesis. A methodology that can unravel different aspects of the Internet infrastructure and performance from the user perspective as well as provide an independent tool for advertisers to measure the quality of their advertising campaigns. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Marco Mellia.- Secretario: Marcelo Gabriel Bagnulo Braun.- Vocal: Andrea Elena Lutu

In the last years indoor localization and applications that use positioning information have attracted a lot of attention from the research community as well as the industry. The achieved accuracy of an indoor localization system is the key to enable certain applications, such as navigation. In a scenario where the system runs exclusively on commodity hardware such as smartphones and even without installing any mobile app in the mobile device, location information may be exploited not only for navigation, but also for the benefit of the network itself or for the investigation of physical behaviors. In indoor areas, communication technologies such as Wireless Fidelity (WiFi) are gaining popularity due to the ever increasing availability and deployment of APs that can be used simultaneously as an infrastructure for networking and positioning. Despite being the most used technology for tracking devices in indoor environments, the research community has focused more intensively on approaches such as the Received Signal Strength Indicator (RSSI) or combining WiFi signals with inertial sensors as found in smartphones. These approaches are error prone, dependent on frequent calibrations, or they require specialized hardware or the user must be interested to run a specific application designed to run on the smartphone. Using only commodity WiFi chipsets, this thesis investigates location systems that are based on Time-of-Flight (ToF) echo technique, and do not require any calibration and user intervention, and it explores ranging and location data to improve network management and infer user behaviour. ToF technique has been modestly exploited for indoor localization, whereas ToF is successfully used in Global Positioning System (GPS) in outdoor environments. The reason is that WiFi ToF measurements, mainly extracted from commodity chipsets, suffer from extensive device-related noise which makes it challenging to differentiate between direct path from non-direct path signal components when estimating the ranges. Existing multipath mitigation techniques tend to fail at identifying the direct path when the devicerelated Gaussian noise is in the same order of magnitude, or larger than the multipath noise. In order to address this challenge, we first propose in this thesis a new method for filtering ranging measurements, extracted from a commodity WiFi chipset, that is better suited for the inherent large noise as found in WiFi radios. Our proposed filter combines statistical learning and robust statistics and it does not require specialized hardware, the intervention of the user, or cumbersome on-site manual calibration. This makes the method we propose as the first contribution of the present work particularly suitable for indoor localization in large-scale deployments using existing legacy WiFi infrastructures. We build and investigate a multi WiFi ToF-based APs localization system, where these filtered timing signals are used as ranging inputs of the multi-lateration problem for positioning. We then deploy and evaluate our system for indoor mobile tracking scenarios in many multipath-rich environments, across multiple testbeds which cover different surfaces and further test it in Microsoft indoor localization competitions, demonstrating that, despite these challenges, our technique can achieve distance accuracy comparable to other approaches proposed by the start of the art but does not share their aforementioned shortcomings. The deployment of a multi-APs positioning system is targeted to specific areas, such as large offices and shopping malls. In order to bring indoor positioning also to homes and small businesses which typically have a single AP, the next step we envision towards a hybrid single-AP positioning system is a deep inspection and interaction of the Fine Time Measurements (FTM), a type of ToF echo technique standardized recently, and Channel State Information (CSI) for ranging and Angle-Of-Arrival (AOA) for angle estimates, respectively. We exploit Physical Layer (PHY) information to detect the number of paths and their directions and we use this information to derive a new method for filtering ranging measurements obtained with the FTM protocol. We achieve sub-meter distance estimation accuracy eliminating the adverse effect of multipath in FTM using calibrated inputs from CSI. We then evaluate the system in multipath-rich environments, demonstrating the capability of the combination of AOA estimation and the proposed FTM refinement approach to achieve reasonable positioning accuracy in areas comparable to typical flat sizes just using a commodity smartphone as target device. All the contributions described so far and all the extracted location data may be exploited also in the network core to better allocate network resources based on the expected link performance. Thus, we take advantage of positioning data and more in general context information, to enable reliable mobile communications via advanced resource management policies and adaptive traffic engineering strategies. Of particular interest for this thesis is to investigate usage of positioning data in industrial environments, where the presence of metallic objects challenge the reliability of wireless communication. With the advent of the fourth Industrial revolution (Industry 4.0) such harsh environments have attracted high attention and, for this reason, of particular interest for this thesis is also the deployment of our multi-APs system in such industrial environment. The latter, due to blockage and strong reflections, is notorious for being adverse to wireless communication, impacting on the signal quality. We propose to exploit the knowledge of location to derive context information to dynamically allocate wireless resources in time and space to target devices. We exploit the spatial geometry of the APs and a statistical model that maps the user position’s spatial distribution to an angle error distribution to derive a hypothesis test to declare if the link is in Line-Of-Sight (LOS) or Non-Line-Of-Sight (NLOS). In order to avoid changes to the client side and operate with a single interface radio, we use the same wireless network both for positioning and scheduling. We experimentally show that context information applied to wireless resources protocol help increasing the network throughput in the aforementioned industrial-like scenario. Finally, taking advantage of ranging information extracted from FTM, we propose to estimate the device movement without any access to physical inertial sensors in the mobile. The idea is to infer the movements of the mobile through radio measurements, a concept we call ”virtual inertial sensors”. We propose a method for estimating the user walking speed and a novel method for the rotation of a mobile device. We evaluate and demonstrate the proposed approaches with experiments, and we compare the rotation method with a possible solution that explores CSI measurements. While FTM works with only one single antenna, it achieves better performance than a CSI based estimator that exploits four antennas and multiple sub-carriers at the AP, but are yet limited by the typical one single WiFi antenna at the smartphone side. This new concept of virtual inertial sensors can be leveraged by location systems and sensing mechanisms to improve localization accuracy, infer user behavior, and design better and more secure communication. In conclusion, in this thesis we experimentally demonstrate that ToF data, extracted from commodity WiFi chipsets, can be used for tracking devices and provide context information in indoor environments without the need for environmental calibration and installations of mobile apps, but only with standard-compliant measurements performed from the WiFi infrastructure. Telematics Engineering Universidad Carlos III de Madrid, Spain pub

Fast advance in the design of 5G cellular networks has motivated a lot of research that addresses challenges given by the explosive growth of traffic burden, the rise of energy consumption constraints, the unprecedentedly high demand for broadband mobile connectivity and guaranteed quality-of-service for end-users. Therefore the appearance of new technologies, system designs and fast network solutions becomes vital to bear such high demand in network infrastructures. In this context, the wireless relay scenario has emerged as a key enabler to deal with such challenges. Having clever and efficient schemes that allow traffic to follow alternative relayed paths rather than direct delivery from producer to consumer stands as a crucial need to be properly integrated on the 5G and beyond networks. Depending on the kind of relay, we envision different relay paradigms: users aiming to relief the traffic burden enable device-to-device relay systems; flexible relaying for dense wireless backhaul systems powered by directional transmissions needs smart relay to boost spatial reuse that minimizes the amount of time needed for traffic readiness; and the possibility of mounting relays on extremely-mobile devices such as drones turns the air space into an unexplored vast amount of possibilities to properly position aerial relays. In this thesis, we present practical optimization tools that leverage the mentioned wireless relay paradigms. We derive optimization frameworks that boost important network metrics such as fair traffic delivery, backhaul traffic readiness or network coverage in current cellular networks. We carefully model network features such as traffic paths, consumed energy, user throughput, transmission directionality or link activation cost, among others. Hence, we approach realistic network infrastructures restricted by technical, physical, flow, or fairness constraints. As unavoidable complex mathematical constraints arise that often turn into an NP-Complete problem, we propose lightweight schemes that work in low-degree polynomial time that are able to provide efficient closeto- optimal solutions, as required in current networks operating at tiny time-scales. The results reported in this thesis show that designing optimization tools that properly identify key opportunities for efficient relay such as best split traffic paths, best directional transmission scheduling or best aerial relay positioning provides very high gains in terms of throughput experience, fast readiness of traffic at the edge nodes or users coverage. Hence, solutions proposed in this thesis comply with implementation requirements as well as guaranteed performance service for desirable integration on current cellular networks. This thesis has been partially supported by the FPU grant (Ayudas para la Formación de Profesorado Universitario) from the Spanish Ministry of Education, Culture and Sports (MECD). Grant reference: FPU2015/02051. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Marco Giuseppe Ajmone Marsan.- Secretario: Paolo Casari.- Vocal: Mehdi Bennis

Future millimeter-wave networks will support very high densities of devices and access points. This vastly increases the overhead required for access point selection and beam training. Due to unfavorable radio propagation, mmWave systems will exploit largescale MIMO and adaptive antenna arrays at both the transmitter and receiver to realize sufficient link margin. Beamforming is vital to overcome the high attenuation in wireless millimeter-wave networks. It enables nodes to steer their antennas in the direction of communication. Fortunately, the quasi-optical properties of millimeter-wave channels make location-based network optimization a highly promising technique to reduce control overhead in such millimeter-wave WLANs. In this thesis we present tools to improve mmWave systems. We start by designing an effective lightweight sector beam-pattern design for using as a baseline for hybrid analog-digital structures. We deal with practical constraints of mmWave transceivers and propose a novel, geometric approach to synthesize multi-beamwidth beam patterns that can be leveraged for simultaneous multi-direction scanning. Then we make use of this multi-direction scanning to create a beam training protocol which effectively accelerates the link establishment by exploiting the ability of mobile users to simultaneously receive from multiple directions. We propose smart beam training and tracking strategies for fast mm-wave link establishment and maintenance under node mobility. We leverage the ability of hybrid analog-digital transceivers to collect channel information from multiple spatial directions simultaneously and formulate a probabilistic optimization problem to model the temporal evolution of the mm-wave channel under mobility. We propose a mechanism to extract full channel state information (CSI) regarding phase and magnitude from coarse signal strength readings on off-the-shelf IEEE 802.11ad devices. Using this CSI, transmitters dynamically compute a transmit beam pattern that maximizes the signal strength at the receiver. Channel properties and antenna design at 60GHz are ideal for path angular information extraction, following an almost ideal geometric channel model. Due to this, we present some localization method specifically designed for the 60GHz band. We merge the ideas presented in this thesis and by extracting channel state information from off-the-shelf routers we estimate the user location to manage a location aware beam-training and device handling method. The resulting scheme can predict blockage, optimize access point association, and select the most suitable antenna beam patterns while significantly reducing the beam training overhead. Programa de Doctorado en Multimedia y Comunicaciones por la Universidad Carlos III de Madrid y la Universidad Rey Juan Carlos Presidente: Edward Knightly.- Secretario: Eva Rajo Iglesias.- Vocal: Sundeep Rangan

Mención Internacional en el título de doctor Visible Light Communication(VLC) has emerged in the last years as a new way to communicate. Using the existing lighting infrastructure, it has great potential to provide high bandwidth and communication security, making it a strong alternative against conventional Radio Frequency (RF) communications. Although VLC addresses several of the problems that RF communications have for specific scenarios, its potential for Internet-of-Things (IoT) applications must still be unleashed. IoT deployments are, by nature, limited in some way. The limitation could be given by the hardware used, as the cost may need to be minimal to have dense realistic deployments; the energy available, which depends on the battery size of the device; and computing power available, which is given by the available energy and the processing power of the device, among others. Therefore, there is an interest in studying the advantages, drawbacks, and limitations of integrating VLC in IoT scenarios with the constraints mentioned above. This is especially necessary in the case of real-life deployments, mainly if the devices used are multi-purpose, and they need to perform other tasks, such as sensing, in addition to communicating. First of all, VLC deployments for IoT use the available dense lighting infrastructure to achieve communication on top of illumination in indoor scenarios. The advantage of such an approach is that it allows reusing the existing infrastructure, improving the coverage, and the energy consumed. Although VLC is energy efficient, it consumes more than just illuminating. If the luminaries are not correctly controlled, energy could be wasted by transmitting from a luminary with little or no effect into the receiver. In DenseVLC, we explore the energy consumption of luminaries in dense deployments, and we propose an approach to optimize the Signal-to-Interference-plus-Noise Ratio (SINR) given an energy budget. In order to do so, we propose to coordinate the transmission done by several independent devices concurrently. We introduce a novel synchronization method that uses the Non Line-Of-Sight (NLOS) component of the signal to tackle this problem. Our approach can improve the average system throughput by 45%, or improve the average power efficiency by 2.3 times, compared to existing solutions. Secondly, even if the required infrastructure follows the design presented above, IoT deployments still need to face one fundamental problem: power management at deployed mobile devices. Having batteries increases the price of the device, its size, the maintenance required, and the ecological impact that the product has. Removing the battery while still being able to operate under realistic circumstances would be desired. In this work, we study what limitations such a system has. We then propose a new communication scheme, combining VLC and RF backscattering, that allows having continuous end-toend communication with a custom-designed battery-free device. We design the hardware, software, and protocol that optimizes each aspect of the system to decrease the power requirement of each component. Finally, we evaluate our system and show that it can run with consumptions as low as 95μW, transmit continuously at 500 bits/second, and achieve more than 20 meters on backscattering distance, even with blockage elements as glass and walls covering the Line-Of-Sight (LOS). Thirdly, we explore one of the multiple applications that the designed VLC systems for IoT allow to implement; device positioning. The majority of the literature requires to have multiple transmitters and/or receivers to achieve localization. The objective of this work is to localize with the minimum amount of necessary hardware, which is critical for IoT applications. We investigate how VLC systems could be used for positioning in dynamic scenarios. Exploiting the fact that, in our scenario, the transmitter and receiver are relatively moving, we propose a mathematical solution that, just using one VLC transmitter and one receiver both equipped with a compass, computes the correct relative position. We then implement our solution in a modified version of OpenVLC and achieve accuracies with less than 5 cm of error. Nevertheless, in this work we assume that the NLOS is non-existant, which may not always be the case. Finally, we try to overcome the problem mentioned above of NLOS reflections for device positioning with a low resource consumption NLOS component detector. Similar work try to solve this problem computing the Channel Impulse Response (CIR), but for systems with limited resources this is impractical because 1) The VLC front-end may not be fast enough for acquiring required data for the CIR calculation or 2) IoT boards are not able to run computationally expensive algorithms in real time. In this thesis, we propose a solution that in complex environments, reduces the localization error using LEDs up to 93%. In order to perform experimental research in VLC for IoT, a research platform is needed. In this thesis, we also present the latest version of an open-source, softwarebased, VLC platform, OpenVLC. OpenVLC was first introduced as part of the thesis of Dr. Qing Wang. During this thesis, the platform has been re-designed on both hardware and software. The throughput improved more than 23 times and the transmission distance increased by a factor of 4. In this thesis, OpenVLC, parts of it, or modified versions have been used as a framework to create new IoT systems and explore the practical side of VLC. As a summary, in this work, we explore how VLC can be leveraged for IoT deployments. We study the features of such real-world deployments from different perspectives in a variety of scenarios, and we show that realistic implementations of VLC systems are not only possible but doable, enabling new features that IoT developer can exploit. This work has been supported by IMDEA Networks Institute, Obra Social Fundación "La Caixa". Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Xinyu Zhang.- Secretario: Vincenzo Mancuso.- Vocal: Julio Francisco Rufo Torres.

De acuerdo con recientes predicciones [1] el tráfico de datos móviles crecerá 7 veces entre 2017 y 2022, llegando a generar 77 exabytes de datos mensuales. Este incremento en el consumo de datos crecerá de la mano de servicios innovadores que incluirán video 8K, realidad virtual, realidad aumentada o juego remoto en la nube, además del gran número de dispositivos inteligentes que formarán parte de las llamadas ciudades inteligentes, casas inteligentes y el internet de las cosas (IoT). Por esta razón, se espera que las redes de transporte móviles del futuro (5G y futuras generaciones) ofrezcan un gran ancho de banda, nunca visto anteriormenente y una latencia mínima que permitan el desarrollo de estos futuros servicios ralmente innovadores. Estas redes del futuro no sólo evolucionarán buscando un incremento sustancial en el rendimiento, también ofrecerán una reducción del gasto de operacion (OPEX) y el gasto de capital (CAPEX) para conseguir un retorno de inversión (ROI) razonable. Estos estrictos requisitos no serán impulsados únicamente por la introducción de especto radioeléctrico de ondas milimétricas, requerirá un completo rediseño de la arquitectura actual, evolucionando hacia una arquitectura basada en servicios. El primer segmento de red que sufrirá este profundo rediseño será la red de acceso (RAN), adoptando el concepto de Centralized/Cloud RAN (C-RAN), separando las distinas funcionalidades de los puntos de acceso, situando las funcionalidades básicas de radio de menor nivel en la capa de protocolos en una unidad remota, centralizando por otro lado las funcionalidades de las capas superiores en unidades centralizadas. Las divisiones funcionales (functional splits) tradicionales utilizadas en redes 4G están basadas en el protocol Common Public Radio Interface (CPRI), este protocolo requiere de una red de circuitos dedicada con unos requisitos de latencia y ancho de banda muy estrictos, estos requisitos de ancho de banda son linealmente proporcionales al número de antenas, lo que prácticamente imposibilita su uso en despliegues 5G ultra densos. Por esa razón nuevos functional splits están siendo investigados, focalizando los esfuerzos en el uso de Ethernet como tecnología convergente, hecho que a su vez difumina la separación entre los segmentos de red de fronthaul y backhaul, conduciendo a una red integrada basada en paquetes, normalmente denominada como crosshaul. Este movimiento hacia una red totalmente paquetizada relaja extremadamente los requisitos de la red de transporte, abriendo oportunidades a soluciones virtualizadas, promoviendo el uso de tecnologías como Software Defined Networking (SDN) y Network Function Virtualization (NFV). SDN ha emergido como una potente herramienta para los operadores para gestionar sus infraestructuras, este paradigma ofrece la posibilidad de ejecutar múltiples aplicaciones inteligentes con el propósito de optimizar la red gracias a la centralización de la lógica de la red en un controlador. Sin embargo, la propia naturaleza que permite esta gestión tan eficiente y flexible, la centralización lógica, a su vez plantea numerosos desafíos debido a la falta de herramienas de monitorización apropiadas para este tipo de arquitecturas de red. Para poder ejecutar decisiones de manera correcta y precisa el controlador (y las aplicaciones que ejecuta) necesita obtener datos y estadístacas de la red de manera continua y actualizada. Sin embargo, esto no es posible con las soluciones de red basadas en SDN existentes actualmente, debido a diversos problemas de escalabilidad y precisión. Además, la centralización lógica del plano de control también presenta problemas de fiabilidad y seguridad ya que un controlador centralizado es un punto único de fallo, siendo un imporante foco de interés para conseguir la fiabilidad requerida en las redes de operador. La clara tendencia hacia la softwarización no se limita solamente a la inclusión del paradigma SDN, va de la mano de la virtualización de funciones de red. Por esta razón las redes móviles del futuro harán un uso intensivo de los centros de datos, ejecutando la mayoría de los servicios de red de manera virtualizada. La ejecución de estos servicios de manera virtualizada no se relizará exclusivamente en grandes centros de datos en la nube (cloud), también se utilizarán pequeños centros de datos situados en el borde de la red (edge) y en la niebla (fog), con el objetivo de reducir la latencia. Esta creciente demanda está llevando a la infraestructura de red de los centros de datos al límite, cuyos requisitos en términos no solo de latencia y ancho de banda, sino también de escalabilidad y fiabilidad cada vez son más estrictos. Los algoritmos de switching tradicionales de centros de datos basados en capa 2 utilizan lookup tables, llenas de direcciones Media Access Control (MAC) de 48-bits donde el switch buscará el siguiente salto al que enviar el paquete. Recientemente, Institute of Electrical and Electronics Engineers (IEEE) 802 ha definido nuevos mecanismos para el uso estructurado del espacio local de direcciones MAC. Esto permite introducir semántica en el espacio de dirección y utilizarlo en aplicaciones para mejorar todo el proceso de reenvío de paquetes. Desde esta perspectiva, esta tesis propone un conjunto de soluciones y arquitecturas basadas en SDN para la siguiente generación de redes, incluyendo tanto el segmento de transporte como los centros de datos. Comenzando con la red de transporte móvil, esta tesis presenta un novedoso diseño de una red crosshaul completamente integrada, implementada y desplegada utilizando whiteboxes. El diseño del pipeline de los elementos de la red crosshaul ha sido cuidadosamente optimizado, elminando todas las operaciones que impactaban negativamente en la latencia, adecuándolo para los functional splits de más bajo nivel. El diseño de la red crosshaul mostrado en esta tesis ha sido cuidadosamente evaluado, convirtiéndose en la primera validación empírica de una red crosshaul completamente integrada, capaz de transportar flujos de diferentes functional splits sobre la misma infraestructura de red. La solución de crosshaul basada en SDN detallada en este trabajo está acompañada de un conjunto de herramientas para mejorar las funciones del plano de control. Esta tesis presenta y evalúa un conjunto de herramientas de monitorización denotadas como Adaptative Telemetry System (ATS), diseñadas con el objetivo de habilitar la ejecución de tareas Operation Administration and Maintenance (OAM) localmente en los switches, soportando la obtención de métricas de red de manera activa y precisa, incluyendo un sistema de reporte al plano de control, cumpliendo con los requisitos exigidos por las redes de operador. La solución propuesta está formada de una aplicación, un plugin y un agente local ATS que permite la configuración remota de procedimientos OAM utilizando máquinas de estado finitas, siendo completamente compatible con el estándar de OpenFlow. Esta tesis también se centra en los problemas de fiabilidad existentes en la propia naturaleza centralizada del paradigma SDN, presentando el diseño y la evaluación de de una solución basada en la utilización de protocolos multipath para robustecer la conectividad del canal de control. La solución presentada en esta tesis utiliza un protocolo multipath para la creación de múltiples caminos redundantes a través de redes in-band y out-of-band, permitiendo migrar la conexión de control de una ruta a otra de manera segura y estable, sin ninguna interrupción. Finalmente, esta tesis también investiga las oportunidades presentes en las nuevos mecanismos de direccionamiento propuestos en IEEE 802 mencionados previamente para su aplicación en redes de centros de datos. Esta tesis incluye un novedoso diseño basado en Software Defined Addressing (SDA) que permite introducir semánticas en el espacio de direccionamiento MAC, dando significado a la direcciones, eliminado por tanto la necesidad de table lookups. Esta tesis detalla una validación unitaria, comparando la solución propuesta con una solución SDN estándar, obteniendo unos resultados realmente positivos. Después de la validación unitaria esta tesis detalla una evaluación completa de un caso de uso potencial de la solución propuesta. Este caso de uso pretende separar los flujos pequeños realmente sensibles a la latencia de los flujos de mayor tamaño que demandan más ancho de banda. El caso de uso propuesto ha sido evaluado en una red de centro de datos emulada, demostrando los beneficios potenciales en térmnos de congestión de red que puede oferecer esta separación de tipos de tráfico. Como conclusión, se puede decir que esta tesis analiza los déficit de las futuras redes móviles, incluyendo tanto las redes de transporte como los centros de datos, proponiendo un conjunto de novedosas soluciones basadas en el paradigma SDN, con el claro objetivo de mejorar la flexibilidad, el rendimiento y la robustez de la red y a su vez reduciendo el coste para cumplir con los requisitos establecidos para los servicios de red del futuro. According to the recent predictions [1] mobile data traffic will increase 7-fold between 2017 and 2022, generating around 77 exabites of data per month. This data traffic increase will be pushed by innovative services including 8K video, virtual reality, augmented reality or cloud gaming, in addition to the numerous devices that will enable smart cities, smart homes, and Internet of Things (IoT). For that reason, future mobile transport networks (5G and beyond) are expected to offer unprecedented bandwidth with extreme low latency to support all this kind of innovative services, but future networks are also expected to reduce the Operational Expenditure (OPEX) and Capital Expenditure (CAPEX) to a reasonable Return on Investment (ROI) range. This stringent requirements will not only be driven by the introduction of the mmWave radio spectrum, it will require a thorough redesign of the current network architecture, moving towards a new service-based mobile system. The first network segment that will undergo this redesign is the Radio Access Network (RAN), adopting the C-RAN concept, decomposing the radio access point functionality into small footprint, placing the basic radio functions of the lower levels of the protocol stack in a remote unit, and centralizing the upper levels of the stack in a Central/Cloud unit. The traditional functional split already utilized in 4G networks, based on CPRI protocol, requires a dedicated circuit-based network with stringent requirements in terms of delay and bandwidth, the bandwidth requirements are linearly dependent with the number of antennas, making this model not feasible for ultra-dense 5G deployments. For that reason new functional splits are being investigated, using Ethernet as the common factor, blurring the traditional separation between the fronthaul and backhaul network segments, driving into a converged packet based network, commonly denoted as Crosshaul. The movement towards new packet-based functional splits relax the requirements in the transport network, enabling new opportunities for softwarized solutions, focusing on SDN and NFV technologies for the future network deployments. SDN has emerged as a basic toolset for operators to manage their infrastructure, as it opens up the possibility of running a multitude of intelligent and advanced applications for network optimization purposes in a centralized network controller. However, the nature that makes possible this efficient management and operation in a flexible way, the logical centralization, also poses important challenges due to the lack of proper monitoring tools suited for SDN-based architectures. In order to take timely and right decisions while operating a network, centralized intelligence applications need to be fed with a continuous stream of up-to-date network statistics. However, this is not feasible with current SDN solutions due to scalability and accuracy issues derived from the centralized intrinsic nature of the SDN paradigm. Furthermore, the control centralization also poses resiliency, reliability and security problems since a centralized controller is a potential single point of failure and a potential bottleneck, needing special attention in order to support the demanded carrier grade requirements. The softwarization trend does not stop with the incorporation of SDN solutions, it goes hand-in-hand with the virtualization of network functions. For that reason, future mobile networks will make extensive use of Data Center Networks (DCNs), running most virtualized network services and data processing. Not only big cloud data centers will store, compute and analyze all the data, also smaller edge and fog data centers will be used, aiming to reduce the overall network latency required in the new generation of mobile networks. This increasing demand is pushing the limits of current DCNs, requiring, like mobile transport networks, higher scalability, resiliency and performance while reducing the network deployment cost. Traditional approaches for layer-2 switching in DCNs rely on table lookups, where a flat 48-bit MAC address space is searched in order to find the next hop to forward the packet to. Very recently, the IEEE 802 has defined new mechanisms for the structured use of the local MAC address space. This new flexibility enables the embedding of semantics in the MAC addresses that can be used for new applications, such as improving the standard layer-2 forwarding process. From this perspective, this thesis proposes a set of SDN-based solutions and architectures for next generation networks, including both mobile transport and DCN. Departing from the mobile transport, this thesis presents a novel design of a fully integrated Crosshaul solution, implemented and deployed using cost-efficient SDN-based forwarding elements. The pipeline design of the Crosshaul forwarding elements presented in this thesis has been carefully optimized, eliminating the latency-impacting operations, making it suitable for lower layer functional splits. The Crosshaul design introduced in this thesis has been extensively evaluated, being the first empirical validation of a fully integrated Crosshaul network, transporting flows from various functional split options over the same switching infrastructure. The presented SDN-based Crosshaul solution is accompanied by a set of tools devoted to enrich the control plane functions. This thesis presents and evaluates a set of monitoring tools denoted as ATS, that enables the local execution of OAM procedures directly on the switches, supporting the active and precise measurements and reporting of strategic metrics, needed for the correct operation of carrier grade networks. The proposed ATS solution is formed by an ATS application, an ATS plugin and an ATS agent, allowing the remote configuration of OAM procedures by using Finite State Machine (FSMs) modelling, being fully compatible with the current OpenFlow standard. This thesis also focuses on the reliability issues present in the SDN centralized nature, presenting the design and evaluation of a multipath-based solution that aims to secure and robustize the control channel. The presented multipath solution creates several disjoint redundant paths over the out-of-band and in-band management networks allowing to seamless migrate the OpenFlow control connection from a failing path to an available one without traffic interruption. Finally, this thesis also investigates the opportunities present in the aforementioned IEEE 802 mechanisms for the structured use of the local MAC address space, and its applicability to current DCNs L2 switching schemas. This thesis includes a novel SDA solution that allows to embed semantics into the MAC addresses, giving meaning to the address structure, eliminating the need of look-up tables. This thesis incorporates a unit testing validation of the proposed SDA solution, comparing it with a baseline switching method. Due to the positive results obtained in the unit testing validation, the presented contribution also evaluates the flexibility of SDA illustrating a potential application for segregating the latency-sensitive small-frame flows from the large data-intensive large flows, evaluating it in an emulated DCN, demonstrating the potential benefits of this traffic segregation in terms of network congestion. As a conclusion, it can be said that this thesis analyzes some technical gaps of the future generation of mobile networks, ranging from mobile transport to data center networks, proposing novel SDN-based solutions aiming to increase the overall flexibility, performance and robustness while reducing the cost in order to match the stringent requirements of future network services. Participation in collaborative European projects: H2020 5G-Crosshaul, H2020 5G-CORAL Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Ignacio Soto Campos.- Secretario: Carlos Donato Morales.- Vocal: Pablo Molinero Fernández

En la actualidad, el modelo de clase tradicional basado en clases magistrales está siendo sustituido cada vez más por otros modelos pedagógicos que ponen el foco del aprendizaje en los estudiantes, en lugar de en el profesor. Uno de estos modelos alternativos es el modelo de clase invertida, que es el modelo de aprendizaje de referencia utilizado en la presente Tesis Doctoral. Debido a las particularidades del modelo de clase invertida, existen diferentes aspectos que pueden ser objeto de mejora, si se dispusiera de información educativa de calidad que permitiera realizar intervenciones basadas en datos objetivos y significativos. Por ejemplo, si se conocieran los conceptos más complejos para los estudiantes, se podría proporcionar ayuda personalizada a los estudiantes que lo necesitaran, o se podría modificar la estructura de la clase presencial para adaptarla a sus necesidades. En este sentido, en la presente Tesis Doctoral se propone el uso de la analítica del aprendizaje (el análisis de los datos educativos de los estudiantes) para la obtención de información educativa con la que poder realizar acciones que permitan mejorar el proceso de aprendizaje de los estudiantes en un contexto educativo que haga uso del modelo de clase invertida. Tras un análisis en profundidad del estado del arte mediante una revisión sistemática del mismo, se ha corroborado la necesidad de investigaciones en relación con el uso de la analítica del aprendizaje junto con la clase invertida. Además, el análisis de los resultados de una serie de entrevistas con gestores, profesores y estudiantes universitarios, ha mostrado necesidades educativas que la propuesta puede llegar a satisfacer. Por tanto, existen motivos justificados para el desarrollo del presente trabajo. En particular, en la presente Tesis Doctoral se define un marco de trabajo teórico en el que se incluye una propuesta de los indicadores asociados a la analítica del aprendizaje más propicios de ser utilizados en diferentes escenarios que hacen uso del modelo de clase invertida, qué información se puede obtener mediante dichos indicadores, y qué intervenciones (acciones) se podrían realizar a través de dicha información de forma que se mejore el proceso de aprendizaje de los estudiantes en un contexto de clase invertida. Por otra parte, con el objetivo de facilitar la implementación práctica del marco de trabajo, se ha desarrollado una herramienta software de analítica del aprendizaje que permite obtener y visualizar, de manera automática, la información descrita en el marco de trabajo. De esta forma, los profesores disponen de toda la información necesaria para la toma de decisiones en un contexto de clase invertida. Además, las visualizaciones implementadas en la herramienta han sido diseñadas considerando las particularidades asociadas al modelo de clase invertida, puesto que muestran información muy relevante en un contexto de clase invertida, de una manera simple de interpretar, y con diferentes configuraciones que pueden ser de utilidad en dicho contexto. Por ejemplo, es posible visualizar la información asociada a diferentes intervalos de tiempo y, además, dichos intervalos se pueden ajustar dependiendo de la estructura del curso correspondiente. No sólo eso, la herramienta también incorpora una funcionalidad para la creación de grupos de estudiantes en base a datos de los propios estudiantes, todo ello en un contexto de clase invertida. En concreto, para crear los grupos se utilizan los datos de los estudiantes asociados a sus interacciones con los vídeos y con los ejercicios que se proporcionan antes de la clase presencial. Dicha funcionalidad tiene el objetivo de facilitar el desarrollo de actividades de aprendizaje colaborativo en escenarios en los que se haga uso del modelo de clase invertida. Finalmente, tanto el marco de trabajo como la herramienta han sido evaluados con el objetivo de determinar su utilidad real. Con este objetivo, se han realizado una serie de experiencias piloto con diferentes profesores en las que se han mostrado tanto el marco de trabajo como la herramienta, tras las cuales se ha realizado una entrevista con cada profesor. Los resultados de dichas experiencias han sido bastante positivos, puesto que los profesores han expresado una opinión favorable tanto del marco de trabajo como de la herramienta. Por otra parte, también se ha evaluado la funcionalidad de la herramienta que permite la creación de grupos de estudiantes. Dicha evaluación ha consistido en el análisis de un conjunto de datos de estudiantes que han seguido el modelo de clase invertida. Los resultados del análisis muestran que la formación de grupos se realiza de manera apropiada y, además, que la herramienta puede ser utilizada, sin perjuicio en sus prestaciones, en escenarios en los que únicamente se disponga de información asociada con la interacción de los estudiantes con vídeos educativos, o en escenarios en los que se disponga exclusivamente de información relacionada con la interacción de los estudiantes con ejercicios online. Palabras clave: clase invertida, analítica del aprendizaje, análisis de datos, visualizaciones, intervenciones, aprendizaje personalizado. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Manuel Alonso Castro Gil.- Secretario: Castro Alario Hoyos.- Vocal: Bernardo Tabuenca Archilla

Decentralized Congestion Control (DCC) mechanisms are a crucial part of Intelligent Transport Systems (ITS) and Vehicular ad-hoc networks (VANETs). It is through the use of DCC that a network maximizes its throughput and efficiency by allocating resources (i.e., access to the medium) fairly in congestion situations. The ITS framework defined by the European Telecommunications Standards Institute (ETSI) considers the usage of DCC both in the Access layer and as part of a cross-layer architecture. This means that DCC mechanisms influence the performance of elements in different layers. One of the components that are directly influenced by DCC is the Cooperative Awareness (CA) basic service —the entity that keeps vehicles aware of the presence of each other on the road. In this thesis, we perform an analysis of the DCC mechanism defined by ETSI, particularly an analytical and experimental evaluation of the performance of the ETSI DCC Adaptive Approach algorithm, comparing it to related algorithms including our own proposal (Dual-alfa). Additionally, we present an evaluation of the effect of DCC on the end-to-end delay of the CA basic service, and propose Generate-on-Time (GoT), an addition to the standard that accounts for delays caused by DCC and the effects these delays bring on other metrics. Results from this thesis show that Dual-alfa is a valid addition to ETSI Adaptive DCC in order to account for transitory situations. Furthermore, the evaluation of Generate-on-Time showed that it improves the performance of Cooperative Awareness Messages (CAMs) when multiple types of traffic coexist. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Pietro Manzoni.- Secretario: María Celeste Campo Vázquez.- Vocal: José Eugenio Naranjo Hernández

Mención Internacional en el título de doctor It is now commonly agreed that future 5G Networks will build upon the network slicing concept. Network slicing is an emerging paradigm in mobile networks that leverages Network Function Virtualization (NFV) to enable the instantiation of multiple logically independent copies -named slices- of a same physical network infrastructure. The operator can allocate to each slice dedicated resources and customized functions that allow meeting the highly heterogeneous and stringent requirements of modern mobile services. Managing functions and resources under network slicing is a challenging task that requires making efficient decisions at all network levels and in real-time, which can be achieved by integrating Artificial Intelligence (AI) in the network. This thesis investigates the potential of AI for sliced mobile networks. In particular it focuses on resource allocation and orchestration for network slices. This involves two steps: (i)Admission Control that is responsible to decide which slices can be admitted to the network, and (ii) Network resource orchestration that dynamically allots to the admitted slices the necessary resources for their operation. Network Slicing will have an impact on the models that sustain the business ecosystem opening the door to new players: the Infrastructure Provider (InP), which is the owner of the infrastructure, and the tenants, which may acquire a network slice from the InP to deliver specific service to their customers. In this new context, how to correctly handle resource allocation among tenants and how to maximize the monetization of the infrastructure become fundamental problems that need to be solved. In this thesis we address this issue by designing a network slice admission control algorithm that (i) autonomously learns the best acceptance policy while (ii) it ensures that the service guarantees provided to tenants are always satisfied. This includes (i) an analytical model for the admissibility region of a network slicing-capable 5G Network, (ii) the analysis of the system (modeled as a Semi-Markov Decision Process) and the optimization of the infrastructure provider’s revenue, and (iii) the design of a machine learning algorithm that can be deployed in practical settings and achieves close to optimal performance. Dynamically orchestrate network resources is both a critical and challenging task in upcoming multi-tenant mobile networks, which requires allocating capacity to individual network slices so as to accommodate future time-varying service demands. Such an anticipatory resource configuration process must be driven by suitable predictors that take into account all the sources of monetary cost associated to network capacity orchestration. Legacy models that aim at forecasting traffic demands fail to capture these key economic aspects of network operation. To close this gap in the second part of this thesis, we first present DeepCog, a first generation deep neural network architecture inspired by advances in image processing and trained via a dedicated loss function in order to deal with monetary cost due to overprovisioning or underprovisioning of networking capacity. Unlike traditional traffic volume predictors, DeepCog returns a cost-aware capacity forecast, which can be directly used by operators to take short- and long-term reallocation decisions that maximize their revenues. Extensive performance evaluations with real-world measurement data collected in a metropolitan-scale operational mobile network demonstrate the effectiveness of our proposed solution, which can reduce resource management costs by over 50% in practical case studies. Then we introduce AZTEC, a second generation data-driven framework that effectively allocates capacity to individual slices by adopting an original multi-timescale forecasting model. Hinging on a combination of Deep Learning architectures and a traditional optimization algorithm, AZTEC anticipates resource assignments that minimize the comprehensive management costs induced by resource overprovisioning, instantiation and reconfiguration, as well as by denied traffic demands. Experiments with real-world mobile data traffic show that AZTEC dynamically adapts to traffic fluctuations, and largely outperforms state-of-the-art solutions for network resource orchestration. At the time of writing DeepCog and AZTEC are, to the best of our knowledge, the only works where a deep learning architecture is explicitly tailored to the problem of anticipatory resource orchestration in mobile networks. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Arturo Azcorra Saloña.- Secretario: Giovanni Neglia.- Vocal: Patrick Thiran

Mención Internacional en el título de doctor The Electromagnetic (EM) spectrum is well regulated by frequency assignment authorities, national regulatory agencies and the International Communication Union (ITU). Nowadays more and more devices such as mobile phones and Internet-of-Things (IoT) sensors make use of wireless communication. Additionally we need a more efficient use and a better understanding of the EM space to allocate and manage efficiently all communications. Governments and telecommunication operators perform spectrum measurements using expensive and bulky equipments scheduling very specific and limited spectrum campaigns. However, this approach does not scale as it can not allow to widely scan and analyze the spectrum 24/7 in real time due to the high cost of the large deployment. A pervasive deployment of spectrum sensors is required to solve this problem, allowing to monitor and analyze the EM radio waves in real time, across all possible frequencies, and physical locations. This thesis presents ElectroSense, a crowdsourcing and collaborative system that enables large scale deployments using Internet-of-Things (IoT) spectrum sensors to collect EM spectrum data which is analyzed in a big data infrastructure. The ElectroSense platform seeks a more efficient, safe and reliable real-time monitoring of the EM space by improving the accessibility and the democratization of spectrum data for the scientific community, stakeholders and the general public. In this work, we first present the ElectroSense architecture, and the design challenges that must be faced to attract a large community of users and all potential stakeholders. It is envisioned that a large number of sensors deployed in ElectroSense will be at affordable cost, supported by more powerful spectrum sensors when possible. Although low-cost Radio Frequency (RF) sensors have an acceptable performance for measuring the EM spectrum, they present several drawbacks (e.g. frequency range, Analog-to-Digital Converter (ADC), maximum sampling rate, etc.) that can negatively affect the quality of collected spectrum data as well as the applications of interest for the community. In order to counteract the above-mentioned limitations, we propose to exploit the fact that a dense network of spectrum sensors will be in range of the same transmitter(s). We envision to exploit this idea to enable smart collaborative algorithms among IoT RF sensors. In this thesis we identify the main research challenges to enable smart collaborative algorithms among low-cost RF sensors. We explore different crowdsourcing and collaborative scenarios where low-cost spectrum sensors work together in a distributed manner. First, we propose a fast and precise frequency offset estimation method for lowcost spectrum receivers that makes use of Long Term Evolution (LTE) signals broadcasted by the base stations. Second, we propose a novel, fast and precise Time-of-Arrival (ToA) estimation method for aircraft signals using low-cost IoT spectrum sensors that can achieve sub-nanosecond precision. Third, we propose a collaborative time division approach among sensors for sensing the spectrum in order to reduce the network uplink bandwidth for each spectrum sensor. By last, we present a methodology to enable the signal reconstruction in the backend. By multiplexing in frequency a certain number of non-coherent low-cost spectrum sensors, we are able to cover a signal bandwidth that would not otherwise be possible using a single receiver. At the time of writing we are the first looking at the problem of collaborative signal reconstruction and decoding using In-phase & Quadrature (I/Q) data received from low-cost RF sensors. Our results reported in this thesis and obtained from the experiments made in real scenarios, suggest that it is feasible to enable collaborative spectrum monitoring strategies and signal decoding using commodity hardware as RF sensing sensors. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Bozidar Radunovic.- Secretario: Paolo Casari.- Vocal: Fco. Javier Escribano Aparicio

Mención Internacional en el título de doctor La inminente 5ª generación de sistemas móviles (5G) está a punto de revolucionar la industria, trayendo una nueva arquitectura orientada a los nuevos mercados verticales y servicios. Debido a esto, el 5G Infrastructure Public Private Partnership (5G-PPP) ha especificado una lista de Indicadores de Rendimiento Clave (KPI) que todo sistema 5G tiene que soportar, por ejemplo incrementar por 1000 el volumen de datos, de 10 a 100 veces m´as dispositivos conectados o consumos energéticos 10 veces inferiores. Con el fin de conseguir estos requisitos, se espera expandir los despligues actuales usando mas Puntos de Acceso (PoA) incrementando así su densidad con múltiples tecnologías inalámbricas. Esta estrategia de despliegue masivo tiene una contrapartida en la eficiencia energética, generando un conflicto con el KPI de reducir por 10 el consumo energético. En este contexto, la comunidad investigadora ha propuesto nuevos paradigmas para alcanzar los requisitos impuestos para los sistemas 5G, siendo materializados en tecnologías como Redes Definidas por Software (SDN) y Virtualización de Funciones de Red (NFV). Estos nuevos paradigmas son el primer paso hacia la softwarización de los despliegues móviles, incorporando nuevos grados de flexibilidad y reconfigurabilidad de la Red de Acceso Radio (RAN). En esta tesis, presentamos primero un análisis detallado y caracterización de las redes móviles softwarizadas. Consideramos el software como la base de la nueva generación de redes celulares y, por lo tanto, analizaremos y caracterizaremos el impacto en la eficiencia energética de estos sistemas. La primera meta de este trabajo es caracterizar las plataformas software disponibles para Radios Definidas por Software (SDR), centrándonos en las dos soluciones principales de código abierto: OpenAirInterface (OAI) y srsLTE. Como resultado, proveemos una metodología para analizar y caracterizar el rendimiento de estas soluciones en función del uso de la CPU, rendimiento de red, compatibilidad y extensibilidad de dicho software. Una vez hemos entendido qué rendimiento podemos esperar de este tipo de soluciones, estudiamos un prototipo SDR construido con aceleración hardware, que emplea una plataformas basada en FPGA. Este prototipo está diseñado para incluir capacidad de ser consciente de la energía, permiento al sistema ser reconfigurado para minimizar la huella energética cuando sea posible. Con el fin de validar el diseño de nuestro sistema, más tarde presentamos una plataforma para caracterizar la energía que será empleada para medir experimentalmente el consumo energético de dispositivos reales. En nuestro enfoque, realizamos dos tipos de análisis: a pequeña escala de tiempo y a gran escala de tiempo. Por lo tanto, para validar nuestro entorno de medidas, caracterizamos a través de análisis numérico los algoritmos para la Adaptación de la Tasa (RA) en IEEE 802.11, para entonces comparar nuestros resultados teóricos con los experimentales. A continuación extendemos nuestro análisis a la plataforma SDR acelerada por hardware previamente mencionada. Nuestros resultados experimentales muestran que nuestra sistema puede en efecto reducir la huella energética reconfigurando el despligue del sistema. Entonces, la escala de tiempos es elevada y presentamos los esquemas para Recursos bajo Demanda (RoD) en despliegues de red ultra-densos. Esta estrategia está basada en apagar/encender dinámicamente los elementos que forman la red con el fin de reducir el total del consumo energético. Por lo tanto, presentamos un modelo analítico en dos sabores, un modelo exacto que predice el comportamiento del sistema con precisión pero con un alto coste computacional y uno simplificado que es más ligero en complejidad mientras que mantiene la precisión. Nuestros resultados muestran que estos esquemas pueden efectivamente mejorar la eficiencia energética de los despliegues y mantener la Calidad de Servicio (QoS). Con el fin de probar la plausibilidad de los esquemas RoD, presentamos un plataforma softwarizada que sigue el paradigma SDN, OFTEN (OpenFlow framework for Traffic Engineering in mobile Network with energy awareness). Nuestro diseño está basado en OpenFlow con funcionalidades para hacerlo consciente de la energía. Finalmente, un prototipo real con esta plataforma es presentando, probando así la plausibilidad de los RoD en despligues reales. The upcoming 5th Generation of mobile systems (5G) is about to revolutionize the industry, bringing a new architecture oriented to new vertical markets and services. Due to this, the 5G-PPP has specified a list of Key Performance Indicator (KPI) that 5G systems need to support e.g. increasing the 1000 times higher data volume, 10 to 100 times more connected devices or 10 times lower power consumption. In order to achieve these requirements, it is expected to expand the current deployments using more Points of Attachment (PoA) by increasing their density and by using multiple wireless technologies. This massive deployment strategy triggers a side effect in the energy efficiency though, generating a conflict with the “10 times lower power consumption” KPI. In this context, the research community has proposed novel paradigms to achieve the imposed requirements for 5G systems, being materialized in technologies such as Software Defined Networking (SDN) and Network Function Virtualization (NFV). These new paradigms are the first step to softwarize the mobile network deployments, enabling new degrees of flexibility and reconfigurability of the Radio Access Network (RAN). In this thesis, we first present a detailed analysis and characterization of softwarized mobile networking. We consider software as a basis for the next generation of cellular networks and hence, we analyze and characterize the impact on the energy efficiency of these systems. The first goal of this work is to characterize the available software platforms for Software Defined Radio (SDR), focusing on the two main open source solutions: OAI and srsLTE. As result, we provide a methodology to analyze and characterize the performance of these solutions in terms of CPU usage, network performance, compatibility and extensibility of the software. Once we have understood the expected performance for such platformsc, we study an SDR prototype built with hardware acceleration, that employs a FPGA based platform. This prototype is designed to include energy-awareness capabilites, allowing the system to be reconfigured to minimize the energy footprint when possible. In order to validate our system design, we later present an energy characterization platform that we will employ to experimentally measure the energy consumption of real devices. In our approach, we perform two kind of analysis: at short time scale and large time scale. Thus, to validate our approach in short time scale and the energy framework, we have characterized though numerical analysis the Rate Adaptation (RA) algorithms in IEEE 802.11, and then compare our theoretical results to the obtained ones through experimentation. Next we extend our analysis to the hardware accelerated SDR prototype previously mentioned. Our experimental results show that our system can indeed reduce the energy footprint reconfiguring the system deployment. Then, the time scale of our analysis is elevated and we present Resource-on-Demand (RoD) schemes for ultradense network deployments. This strategy is based on dynamically switch on/off the elements that form the network to reduce the overall energy consumption. Hence, we present a analytic model in two flavors, an exact model that accurately predicts the system behaviour but high computational cost and a simplified one that is lighter in complexity while keeping the accuracy. Our results show that these schemes can effectively enhance the energy efficiency of the deployments and mantaining the Quality of Service (QoS). In order to prove the feasibility of RoD, we present a softwarized platform that follows the SDN paradigm, the OFTEN (Open Flow framework for Traffic Engineering in mobile Networks with energy awareness) framework. Our design is based on OpenFlow with energy-awareness functionalities. Finally, a real prototype of this framework is presented, proving the feasibility of the RoD in real deployments. FP7-CROWD (2013-2015) CROWD (Connectivity management for eneRgy Optimised Wireless Dense networks).-- H2020-Flex5GWare (2015-2017) Flex5GWare (Flexible and efficient hardware/software platforms for 5G network elements and devices). Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Gramaglia , Marco.- Secretario: José Nuñez.- Vocal: Fabrizio Giuliano

One of the most relevant issues in the new Digital Transformation is how telcos could create new 5G ecosystems. The use of new 5G technologies enters the telecommunication market in a very disruptive way. Telcos could be prepared for another technological change since the telco industry lived many transformations in the past. With 5G and their disruptive technologies, the result will be, making telcos enter in a non-telco area. Because of that, they will need to interact with different verticals and new partners. There are many technical kinds of literature covering these new 5G technologies, but few of them, cover how telcos should manage and do business with them. Some studies affirm that The telco companies who do not understand that they must interact with others as part of the new ecosystem will eventually die. However, not all are bad news for telcos, the future is not written. There are new opportunities and considering the logical technological restrictions and the possible and realistic possibilities, there is an open space that will be defined by the decisions that will be taken by the industry. This paper aims to clarify these concepts behind the new business models which are its usages, and its roles in the information system domain. To do that, the paper identifies the terminologies used to describe new business models and reuses the previous literature to elaborate on the research. General usages, roles, and potential of the concept are also outlined. The intention of this dissertation is to be used by non-technical people. Basic concepts are explained in a very simple way to allow readers to have an overview of the technical issues needed to understand the conclusions. How to present technical information, in a non-technical way, represents an extra challenge for the author. Finally, concrete and pragmatic proposals will be offered to telco managers and CEOs, to start working in the direction towards monetization of 5G, helping them with company decisions. Master in NFV and SDN for 5G Networks. Curso 2018/2019

Mención Internacional en el título de doctor Millimeter wave (mmWave) technology is one of the main pillars of the next generation Wireless Local Area Networks (WLANs) and 5G mobile networks. The main reason lies in the quantum leap of capacity it provides with respect to wireless networks operating in the sub6-GHz band. Nevertheless, efficient and reliable communication in the mmWave band demands novel techniques to tackle all the barriers associated with wireless propagation in this band. For example, material penetration and diffraction in the mmWave band are much weaker than the ones experienced in the sub-6-GHz band. As a result, wireless propagation in the mmWave band has a quasi-optical behavior in which the Line-of-sight (LoS) component contributes to the majority of the received signal power. For these reasons, mmWave devices rely on either horn antennas or electronically steerable phased antenna arrays to establish directional beams and focus the energy towards a specific direction in space. Although directional communication compensates for the high power attenuation, it creates a new problem that is uncharted for wireless networks operating in the sub-6-GHz band. It makes mmWave wireless links susceptible to blockage, human mobility, and device rotation. Solving these issues requires developing algorithms to quickly find alternative paths for communication upon link interruption. The IEEE 802.11ad protocol is the first WLAN standard that supports wireless networking in the unlicensed 60 GHz band. IEEE 802.11ad tackles the aforementioned problems by introducing new mechanisms at the medium access control (MAC) and Physical (PHY) layers such as beamforming training and beam tracking, hybrid channel access scheme, relay operation mode, multi-band operation, etc. The performance of 60 GHz networks is the result of the interaction of all layers of the protocol stack with the wireless medium. To understand this interaction, it is fundamental to consider 60 GHz networks as a whole. Nevertheless, real-world experimentation with mmWave communication is not always feasible due to the significant amount of resources required and its associated costs. For these reasons, we develop in this thesis a high fidelity system-level model to simulate the IEEE 802.11ad standard. This allows us to study large-scale wireless networks operating in the 60 GHz band, taking into account all of the essential features supported by the standard. We investigate the networking aspects of various mmWave wireless network deployments and provide solutions to boost their performance. Additionally, since most of the mmWave devices are anticipated to adopt the carrier sense multiple access with collision avoidance (CSMA/CA) as the primary channel access scheme, we propose two frame-aggregation policies that significantly improve network throughput and reduce end-to-end delay. To complement our insights from the simulations, we analyze in-depth the performance of Commercial off-the-shelf (COTS) devices that implement the full 802.11ad protocol stack. Fortunately, a growing number of 60 GHz devices supporting the IEEE 802.11ad standard have become recently available. However, the standard does not specify implementation-dependent characteristics that have a significant impact on device performance. For example, this includes the periodicity of beamforming training, which directly affects ongoing communication and performance under mobility. Also, the placement of the 60 GHz antenna in the device plays a fundamental role regarding self-shadowing, which in turn affects how the device shall be deployed to maximize coverage and improve spatial reuse. Understanding such implementation-dependent issues is crucial to correctly draw the aforementioned system-level insights. In this thesis, we characterize and compare commercial 60 GHz devices which are widely used for research purposes. Particularly, we look at different networking aspects, including spatial sharing, beam patterns synthesis, frame aggregation, and the interactions between CSMA/CA and Transmission Control Protocol (TCP) protocols within dense network settings. In summary, we find significant discrepancies in terms of behavior and performance between these devices and what theory suggests. Additionally, achieving a gigabit of throughput and maintaining low-latency require adopting cross-layer solutions operating between the MAC layer and the transport layer. The firmware running on these devices provide some experimental capabilities based on the operation of the IEEE 802.11ad protocol. These capabilities include reporting the Channel State Information (CSI) per antenna element for the connected antenna array. This CSI information contains valuable information about the spatial environment. Additionally, this information is sensitive to any minor changes in the environment. This inspires us to build a collaborative scheme consisting of spatially distributed access points (APs) that sense the environment and pinpoint the location of an obstacle without involving the end users. In this thesis, we first delve into the operation of the IEEE 802.11ad protocol and analyze its performance both in simulations and practice. Second, we propose a set of solutions and recommendations to boost its efficiency. Finally, we extend its scope beyond typical wireless communication and utilize it to build a passive localization system. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Jeroen Maurice Margareta Famaey.- Secretario: Vincenzo Mancuso.- Vocal: Danilo de Donno

Mención Interancional en el título de doctor The architecture of current operator infrastructures is being challenged by the non-stop growing demand of data hungry services appearing every day. While currently deployed operator networks have been able to cope with traffic demands so far, the architectures for the 5th generation of mobile networks (5G) are expected to support unprecedented traffic loads while decreasing costs associated with the network deployment and operations. Indeed, the forthcoming set of 5G standards will bring programmability and flexibility to levels never seen before. This has required introducing changes in the architecture of mobile networks, enabling different features such as the split of control and data planes, as required to support rapid programming of heterogeneous data planes. Network softwarisation is hence seen as a key enabler to cope with such network evolution, as it permits controlling all networking functions through (re)programming, thus providing higher flexibility to meet heterogeneous requirements while keeping deployment and operational costs low. A great diversity in terms of traffic patterns, multi-tenancy, heterogeneous and stringent traffic requirements is therefore expected in 5G networks. Software Defined Networking (SDN) and Network Function Virtualisation (NFV) have emerged as a basic tool-set for operators to manage their infrastructure with increased flexibility and reduced costs. As a result, new 5G services can now be envisioned and quickly programmed and provisioned in response to user and market necessities, imposing a paradigm shift in the services design. However, such flexibility requires the 5G transport network to undergo a profound transformation, evolving from a static connectivity substrate into a service-oriented infrastructure capable of accommodating the various 5G services, including Ultra-Reliable and Low Latency Communications (URLLC). Moreover, to achieve the desired flexibility and cost reduction, one promising approach is to leverage virtualisation technologies to dynamically host contents, services, and applications closer to the users so as to offload the core network and reduce the communication delay. This thesis tackles the above challengeswhicharedetailedinthefollowing. A common characteristic of the 5G servicesistheubiquityandthealmostpermanent connection that is required from the mobile network. This really imposes a challenge in thesignallingproceduresprovidedtogettrack of the users and to guarantee session continuity. The mobility management mechanisms will hence play a central role in the 5G networks because of the always-on connectivity demand. Distributed Mobility Management (DMM) helps going towards this direction, by flattening the network, hence improving its scalability,andenablinglocalaccesstotheInternet and other communication services, like mobile-edge clouds. Simultaneously, SDN opens up the possibility of running a multitude of intelligent and advanced applications for network optimisation purposes in a centralised network controller. The combination of DMM architectural principles with SDN management appears as a powerful tool for operators to cope with the management and data burden expected in 5G networks. To meet the future mobile user demand at a reduced cost, operators are also looking at solutions such as C-RAN and different functional splits to decrease the cost of deploying and maintaining cell sites. The increasing stress on mobile radio access performance in a context of declining revenues for operators is hence requiring the evolution of backhaul and fronthaul transport networks, which currently work decoupled. The heterogeneity of the nodes and transmisión technologies inter-connecting the fronthaul and backhaul segments makes the network quite complex, costly and inefficient to manage flexibly and dynamically. Indeed, the use of heterogeneous technologies forces operators to manage two physically separated networks, one for backhaul and one forfronthaul. In order to meet 5G requirements in a costeffective manner, a unified 5G transport network that unifies the data, control, and management planes is hence required. Such an integrated fronthaul/backhaul transport network, denoted as crosshaul, will hence carry both fronthaul and backhaul traffic operating over heterogeneous data plane technologies, which are software-controlled so as to adapt to the fluctuating capacity demand of the 5G air interfaces. Moreover, 5G transport networks will need to accommodate a wide spectrum of services on top of the same physical infrastructure. To that end, network slicing is seen as a suitable candidate for providing the necessary Quality of Service (QoS). Traffic differentiation is usually enforced at the border of the network in order to ensure a proper forwarding of the traffic according to its class through the backbone. With network slicing, the traffic may now traverse many slice edges where the traffic policy needs to be enforced, discriminated and ensured, according to the service and tenants needs. However, the very basic nature that makes this efficient management and operation possible in a flexible way – the logical centralisation – poses important challenges due to the lack of proper monitoring tools, suited for SDN-based architectures. In order to take timely and right decisions while operating a network, centralised intelligence applications need to be fed with a continuous stream of up-to-date network statistics. However, this is not feasible with current SDN solutions due to scalability and accuracy issues. Therefore, an adaptive telemetry system is required so as to support the diversity of 5G services and their stringent traffic requirements. The path towards 5G wireless networks alsopresentsacleartrendofcarryingoutcomputations close to end users. Indeed, pushing contents, applications, and network functios closer to end users is necessary to cope with thehugedatavolumeandlowlatencyrequired in future 5G networks. Edge and fog frameworks have emerged recently to address this challenge. Whilst the edge framework was more infrastructure-focused and more mobile operator-oriented, the fog was more pervasive and included any node (stationary or mobile), including terminal devices. By further utilising pervasive computational resources in proximity to users, edge and fog can be merged to construct a computing platform, which can also be used as a common stage for multiple radio access technologies (RATs) to share their information, hence opening a new dimension of multi-RAT integration. La arquitectura de las infraestructuras actuales de los operadores está siendo desafiada por la demanda creciente e incesante de servicios con un elevado consumo de datos que aparecen todos los días. Mientras que las redes de operadores implementadas actualmente han sido capaces de lidiar con las demandas de tráfico hasta ahora, se espera que las arquitecturas de la quinta generación de redes móviles (5G) soporten cargas de tráfico sin precedentes a la vez que disminuyen los costes asociados a la implementación y operaciones de la red. De hecho, el próximo conjunto de estándares 5G traerá la programabilidad y flexibilidad a niveles nunca antes vistos. Esto ha requerido la introducción de cambios en la arquitectura de las redes móviles, lo que permite diferentes funciones, como la división de los planos de control y de datos, según sea necesario para soportar una programación rápida de planos de datos heterogéneos. La softwarisación de red se considera una herramienta clave para hacer frente a dicha evolución de red, ya que proporciona la capacidad de controlar todas las funciones de red mediante (re)programación, proporcionando así una mayor flexibilidad para cumplir requisitos heterogéneos mientras se mantienen bajos los costes operativos y de implementación. Por lo tanto, se espera una gran diversidad en términos de patrones de tráfico, multi-tenancy, requisitos de tráfico heterogéneos y estrictos en las redes 5G. Software Defined Networking (SDN) y Network Function Virtualisation (NFV) se han convertido en un conjunto de herramientas básicas para que los operadores administren su infraestructura con mayor flexibilidad y menores costes. Como resultado, los nuevos servicios 5G ahora pueden planificarse, programarse y aprovisionarse rápidamente en respuesta a las necesidades de los usuarios y del mercado, imponiendo un cambio de paradigma en el diseño de los servicios. Sin embargo, dicha flexibilidad requiere que la red de transporte 5G experimente una transformación profunda, que evoluciona de un sustrato de conectividad estática a una infraestructura orientada a servicios capaz de acomodar los diversos servicios 5G, incluso Ultra-Reliable and Low Latency Communications (URLLC). Además, para lograr la flexibilidad y la reducción de costes deseadas, un enfoque prometedores aprovechar las tecnologías de virtualización para alojar dinámicamente los contenidos, servicios y aplicaciones más cerca de los usuarios para descargar la red central y reducir la latencia. Esta tesis aborda los desafíos anteriores que se detallan a continuación. Una característica común de los servicios 5G es la ubicuidad y la conexión casi permanente que se requiere para la red móvil. Esto impone un desafío en los procedimientos de señalización proporcionados para hacer un seguimiento de los usuarios y garantizar la continuidad de la sesión. Por lo tanto, los mecanismos de gestión de la movilidad desempeñarán un papel central en las redes 5G debido a la demanda de conectividad siempre activa. Distributed Mobility Management (DMM) ayuda a ir en esta dirección, al aplanar la red, lo que mejora su escalabilidad y permite el acceso local a Internet y a otros servicios de comunicaciones, como recursos en “nubes” situadas en el borde de la red móvil. Al mismo tiempo, SDN abre la posibilidad de ejecutar una multitud de aplicaciones inteligentes y avanzadas para optimizar la red en un controlador de red centralizado. La combinación de los principios arquitectónicos DMM con SDN aparece como una poderosa herramienta para que los operadores puedan hacer frente a la carga de administración y datos que se espera en las redes 5G. Para satisfacer la demanda futura de usuarios móviles a un coste reducido, los operadores también están buscando soluciones tales como C-RAN y diferentes divisiones funcionales para disminuir el coste de implementación y mantenimiento de emplazamientos celulares. El creciente estrés en el rendimiento del acceso a la radio móvil en un contexto de menores ingresos para los operadores requiere, por lo tanto, la evolución de las redes de transporte de backhaul y fronthaul, que actualmente funcionan disociadas. La heterogeneidad de los nodos y las tecnologías de transmisión que interconectan los segmentos de fronthaul y backhaul hacen que la red sea bastante compleja, costosa e ineficiente para gestionar de manera flexible y dinámica. De hecho, el uso de tecnologías heterogéneas obliga a los operadores a gestionar dos redes separadas físicamente, una para la red de backhaul y otra para el fronthaul. Para cumplir con los requisitos de 5G de manera rentable, se requiere una red de transporte única 5G que unifique los planos de control, datos y de gestión. Dicha red de transporte fronthaul/backhaul integrada, denominada “crosshaul”, transportará tráfico de fronthaul y backhaul operando sobre tecnologías heterogéneas de plano de datos, que están controladas por software para adaptarse a la demanda de capacidad fluctuante de las interfaces radio 5G. Además, las redes de transporte 5G necesitarán acomodar un amplio espectro de servicios sobre la misma infraestructura física y el network slicing se considera un candidato adecuado para proporcionar la calidad de servicio necesario. La diferenciación del tráfico generalmente se aplica en el borde de la red para garantizar un reenvío adecuado del tráfico según su clase a través de la red troncal. Con el networkslicing, el tráfico ahora puede atravesar muchos fronteras entre “network slices” donde la política de tráfico debe aplicarse, discriminarse y garantizarse, de acuerdo con las necesidades del servicio y de los usuarios. Sin embargo, el principio básico que hace posible esta gestión y operación eficientes de forma flexible – la centralización lógica – plantea importantes desafíos debido a la falta de herramientas de supervisión necesarias para las arquitecturas basadas en SDN. Para tomar decisiones oportunas y correctas mientras se opera una red, las aplicaciones de inteligencia centralizada necesitan alimentarse con un flujo continuo de estadísticas de red actualizadas. Sin embargo, esto no es factible con las soluciones SDN actuales debido a problemas de escalabilidad y falta de precisión. Por lo tanto, se requiere un sistema de telemetría adaptable para respaldar la diversidad de los servicios 5G y sus estrictos requisitos de tráfico. El camino hacia las redes inalámbricas 5G también presenta una tendencia clara de realizar acciones cerca de los usuarios finales. De hecho, acercar los contenidos, las aplicaciones y las funciones de red a los usuarios finales es necesario para hacer frente al enorme volumen de datos y la baja latencia requerida en las futuras redes 5G. Los paradigmas de “edge” y “fog” han surgido recientemente para abordar este desafío. Mientras que el edge está más centrado en la infraestructura y más orientado al operador móvil, el fog es más ubicuo e incluye cualquier nodo (fijo o móvil), incluidos los dispositivos finales. Al utilizar recursos de computación de propósito general en las proximidades de los usuarios, el edge y el fog pueden combinarse para construir una plataforma de computación, que también se puede utilizar para compartir información entre múltiples tecnologías de acceso radio (RAT) y, por lo tanto, abre una nueva dimensión de la integración multi-RAT. Programa Oficial de Doctorado en Ingeniería Telemática Presidente: Carla Fabiana Chiasserini.- Secretario: Vincenzo Mancuso.- Vocal: Diego Rafael López García

Millimeter-wave (mmWave) networks are one of the main key components in next cellular and WLANs (Wireless Local Area Networks). mmWave networks are capable of providing multi gigabit-per-second rates with very directional low-interference and high spatial reuse links. In 2013, the first 60 GHz wireless solution for WLAN appeared in the market. These were wireless docking stations under theWiGig protocol. Today, in 2019, 60 GHz communications have gained importance with the IEEE 802.11ad amendment with different products on the market, including routers, laptops and wireless Ethernet solutions. More importantly, mmWave networks are going to be used in next generation cellular networks, where smartphones will be using the 28 GHz band. For backbone links, 60 GHz communications have been proposed due to its higher directionality and unlicensed use. This thesis fits in this frame of constant development of themmWave bands to meet the needs of latency and throughput that will be necessary to support future communications. In this thesis, we first characterize the cost-effective design of COTS (commercial off-the-shelf) 60 GHz devices and later we improve their two main weaknesses, which are their low link distance and their non-ideal spatial reuse. It is critical to take into consideration the cost-effective design of COTS devices when designing networking mechanisms. This is why in this thesis we do the first-of-its-kind COTS analysis of 60 GHz devices, studying the D5000 WiGig Docking station and the TP-Link Talon IEEE 802.11ad router. We include static measurements such as the synthesized beam patterns of these devices or an analysis of the area-wide coverage that these devices can fulfill. We perform a spatial reuse analysis and study the performance of these devices under user mobility, showing how robust the link can be under user movement. We also study the feasibility of having flying mmWave links. We mount a 60 GHz COTS device into a drone and perform different measurement campaigns. In this first analysis, we see that these 60 GHz devices have a large performance gap for the achieved communication range as well as a very low spatial reuse. However, they are still suitable for low density WLANs and for next generation aerial micro cell stations. Seeing that these COTS devices are not as directional as literature suggests, we analyze how channels are not as frequency stable as expected due to the large amount of reflected signals. Ideally, frequency selective techniques could be used in these frequency selective channels in order to enlarge the range of these 60 GHz devices. To validate this, we measure real-world 60 GHz indoor channels with a bandwidth of 2 GHz and study their behavior with respect to techniques such as bitloading, subcarrier switch-off, and waterfilling. To this end, we consider a Orthogonal Frequency-Division Multiplexing (OFDM) channel as defined in the IEEE 802.11ad standard and show that in point of fact, these techniques are highly beneficial in mmWave networks allowing for a range extension of up to 50%, equivalent to power savings of up to 7 dB. In order to increase the very limited spatial reuse of these wireless networks, we propose a centralized system that allows the network to carry out the beam training process not only to maximize power but also taking into account other stations in order to minimize interference. This system is designed to work with unmodified clients. We implement and validate our system on commercial off-the-shelf IEEE 802.11ad hardware, achieving an average throughput gain of 24.67% for TCP traffic, and up to a twofold throughput gain in specific cases. Programa de Doctorado en Multimedia y Comunicaciones por la Universidad Carlos III de Madrid y la Universidad Rey Juan Carlos Presidente: Andrés García Saavedra.- Secretario: Matilde Pilar Sánchez Fernández.- Vocal: Ljiljana Simic

Mención Internacional en el título de doctor Internet Service Providers (ISP) are eagerly looking for obtaining metadata from the traffic that they carry. The obtained metadata is a valuable asset for ISPs to enhance their functionality and reduce their operational cost. Classifying a network traffic based on the application (app) that generates the traffic is vital for today’s ISPs and network providers. They use Network Traffic Classification (NTC) to improve many aspects of their network like security and resource allocation. In addition, NTC enables the ISPs to offer new services to their customers and end users. However, NTC faces a big challenge due to the high dynamic Internet ecosystem. Thousands apps are published daily[1] and NTC needs to be updated with their footprint. Moreover some of the existing apps do not follow IANA[2] port number assignment list to use port number which provides more complexity to the ecosystem. Besides, encryption is a trend to secure end-to-end communication and it makes performing NTC hard for those classifiers who relay on information in users payload. Last but not least the volume of traffic that NTC has to investigate is drastically increasing. Therefore, NTC should be fast enough to do the classification on-line which is an essential requirement for many NTC applications. in this thesis, I propose Chain as a novel algorithm to do NTC. Chain sequentially investigates different aspects of a network traffic and brings a significant improvement in tradeoff between classification performance and speed. Besides, it shows a great flexibility to adopt to the new network traffic due to its modularity design. I have implemented Chain in Traffic Identification Engine (TIE) [3] platform and have evaluated its performance with data set [4] which is published by CBA research group at Technical University of Catalunya. Following I have developed a platform named GTEngin to collect ground truth driven from mobile apps and then I have reevaluated the performance of my proposal with the new ground truth. In addition, I participated in an investigation carrying out on mobile Internet to study the possibility of improving my proposal performance in mobile ecosystem.Consequently, I leverage the result of the investigation and measure the enhancement of my proposal performance which achieved accordingly. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Carlos García Rubio.- Secretario: Francisco Javier Simó Reigadas.- Vocal: Roberto Bifulco

As the Internet is becoming a fundamental aspect of the society serving as a de facto platform for social and business activities, traditional offline activities and services have remarkably migrated to the web. The presence and interaction of users on these platforms has created a large amount of digital trace which is being effectively exploited by businesses targeting users on these platforms. One of the main players in this regard is online advertising which is the underneath business that drives the majority of the most important online services such as social media, search engines, map services, etc. This has made online advertising a crucial Internet service in its own right. While the benefit of using these digital resources has been accepted widely pushing governments and organizations to improve their Internet coverage, there are major challenges that limit the society from enjoying their benefits. Together with transparency and privacy, digital inequality is the main challenge that the society faces today. In this thesis we propose a set of inexpensive and large scale methodologies that leverages datasets from online advertising systems to measure and characterize digital inequality on the web. Our methodologies consider various demographic, geographic and interest categories at global scale that advances the knowledge of the scientific community to better understand the challenges in the interplay between online services and users, specifically digital inequalities. In particular we present three main contributions in this context: (I) A methodology to measure the price variability assigned to users by the online advertising system. We created an advertising price comparison system that leverages the bidding data from four online advertising platforms that contributes to the transparency efforts to understand the economic value that online advertising system assigns to user profiles. Using this data we show that advertising price assigned to user profiles varies depending on the profile of the targeted user. (II) A methodology to leverage social media data for gender based digital inequality research. Efforts to understand global prevalence of gender based digital inequality and its interplay with socioeconomic inequalities are limited by lack of large scale representative dataset. Towards solving this problem we developed inexpensive methodologies using the Facebook online advertising platform to quantify the extent of digital inequality in access to social media and its relation with existing inequality indicators. (III) A methodology to measure and characterize user representation and growth variability on Facebook. Facebook is the most widely used social media platform connecting billions of people globally. However, little is known about composition and growth dynamic of this advertising driven social media giant. The main challenge in measuring these phenomena is due to lack of large scale representative dataset that describe the actual number of users and their activities on the social media. We address this problem by leveraging its online advertising platform to measure and characterize its global composition and growth based on age, gender, and country of location. In summary, the work presented in this thesis contributes to advance our knowledge of digital inequalities measured through social media and to develop methodologies for understanding important socioeconomic issues based on digital traces from social media and the Internet in general. It will motivate further research in the area to understand the extent and causes of digital inequalities in various aspects of the Internet considering different groups of societies. The reported findings and methodologies will also help lay foundation for informed policy development and research to close the gender gap and reduce digital inequalities worldwide. This work has been supported by IMDEA Networks Institute Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Arturo Azcorra Saloña.- Secretario: Carlos Castillo Ocaranza.- Vocal: Gareth Tyson

Mención Internacional en el título de doctor Lately, Artificial Intelligence and Machine Learning (ML) have become game-changing technologies due to their ability to generalize from data and infer algorithmic behaviors that consider larger casuistic that humans are able to. In short, these technologies pursue the installation of human-like intelligence to computer tasks so they can overtake different functions. Despite, their implantation and development in many fields is still too early stage, not to mention the requirements and needs they entail. Therefore, the aim of this thesis is to advance in the application of these technologies and for that we will consider an specific field: The Internet Infrastructure. To this aim, contributions focus on two main specific areas, namely cybersecurity and optical WDM networks. On the security side, we propose a new approach for malware detection and application quality assessment that relies in application meta-information, that is, the data describing the application (such as description, category, permissions...) instead of application code. This approach is detailed and validated in two specific applications: ML-based detection of malware and scalable repackaging detection through meta-data semantic clustering. The first application consists on the usage of meta-data as Machine Learning features with a labeled collection of malware applications to detect whether they are malware or not. Resulting algorithms are capable of detecting malware to a good extent in certain conditions, reaching F-score values of nearly 0.9. Arising from the observations from Machine Learning analysis, Antivirus (AV) engines coming from multi-scanner tools are inspected using data analytics and AI technologies aiming at the understanding of their lack of consensus at the detection and categorization levels. The main aim for this study is twofold: advancing on the understanding of AV detection patterns and policies and the improvement multiengine detection by proposing different aggregation and cleaning tools. Initially, AV engine detections are inspected, showing that most engines disagree when detecting malware to the extent of not completely agreeing in the detection of a single application. Moreover, different detection patterns are observed, namely leader, follower and eccentric engines. At the end, an estimation of the risk of malware per application based on Structural Equation models is proposed. On the family side, we propose a lightweight categorization scheme that achieves comparable scores to other alternatives in the literature at a smaller train cost: SignatureMiner. Using such system, we normalize and categorize AV signatures into 41 distinct families and three broader categories, namely adware, harmful and unknown. Then, an ML classifier to assign and specific category to unknown malware is proposed with high performance. Another application explored for meta-data is that of repackaging detection. Using similarity clustering, a large collection of unlabeled applications from Google Play are inspected and compared to detect potential repackaged applications and their victims. This approach is capable to unveil nearly 420K applications potentially cloned within the Google Play application market. On the network side, we contribute to the introduction of Machine Learning in the field by proposing an integral pipeline framework that improves the development of ML-powered network protocols as enhanced heuristics that emulate optimal solutions in many areas. Such framework is based on data generation, modeling and validation and network implementation. In this thesis, we focus on the first two steps by developing proof of concept solutions for both. Dataset generation and data labeling is addressed with Netgen, a versatile network data generator based on Net2Plan. Netgen functionality is presented and performance and abilities demonstrated. Finally, this thesis addresses the modeling of Routing and Wavelength Assignment (RWA) in its ILP version as an ML problem. The assumption is that ML can be useful to develop an ML-powered heuristic for RWA that performs better than regular heuristics and much faster than ILP and heuristics. Results support the viability of this approach, opening the scheme for other complex network protocols. In sum, this thesis builds different AI-based components to enhance the functionalities and capabilities of different elements in the proposed fields, defining systematic approaches and methodologies to this aim. That way, all works in this document contribute to the design and development of the concept of AI as a Service (AIaaS), that proposes a paradigm for the integration of AI technologies over specific knowledge areas with limited expertise in both AI and the specific area. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Andrés Marín López.- Secretario: Ignacio de Miguel Jiménez.- Vocal: Marco Ruffini

Mención Internacional en el título de doctor Middleboxes, such as proxies, firewalls and NATs play an important role in the modern Internet ecosystem. On one hand, they perform advanced functions, e.g. traffic shaping, security or enhancing application performance. On the other hand, they turn the Internet into a hostile ecosystem for innovation, as they limit the deviation from deployed protocols. It is therefore essential, when designing a new protocol, to first understand its interaction with the elements of the path. The emerging area of crowdsourcing solutions can help to shed light on this issue. Such approach allows us to reach large and different sets of users and also different types of devices and networks to perform Internet measurements. In this thesis, we show how to make informed protocol design choices by expanding the traditional crowdsourcing focus from the human element and using crowdsourcing large scale measurement platforms. We consider specific use cases, namely the case of pervasive encryption in the modern Internet, TCP Fast Open and ECN++. We consider such use cases to advance the global understanding on whether wide adoption of encryption is possible in today’s Internet or the adoption of encryption is necessary to guarantee the proper functioning of HTTP/2. We target ECN and particularly ECN++, given its succession of deployment problems. We then measured ECN deployment over mobile as well as fixed networks. In the process, we discovered some bad news for the base ECN protocol—more than half the mobile carriers we tested wipe the ECN field at the first upstream hop. This thesis also reports the good news that, wherever ECN gets through, we found no deployment problems for the ECN++ enhancement. The thesis includes the results of other more in-depth tests to check whether servers that claim to support ECN, actually respond correctly to explicit congestion feedback, including some surprising congestion behaviour unrelated to ECN. This thesis also explores the possible causes that ossify the modern Internet and make difficult the advancement of the innovation. Network Address Translators (NATs) are a commonplace in the Internet nowadays. It is fair to say that most of the residential and mobile users are connected to the Internet through one or more NATs. As any other technology, NAT presents upsides and downsides. Probably the most acknowledged downside of the NAT technology is that it introduces additional difficulties for some applications such as peer-to-peer applications, gaming and others to function properly. This is partially due to the nature of the NAT technology but also due to the diversity of behaviors of the different NAT implementations deployed in the Internet. Understanding the properties of the currently deployed NAT base provides useful input for application and protocol developers regarding what to expect when deploying new application in the Internet. We develop NATwatcher, a tool to test NAT boxes using a crowdsourcingbased measurement methodology. We also perform large scale active measurement campaigns to detect CGNs in fixed broadband networks using NAT Revelio, a tool we have developed and validated. Revelio enables us to actively determine from within residential networks the type of upstream network address translation, namely NAT at the home gateway (customer-grade NAT) or NAT in the ISP (Carrier Grade NAT). We deploy Revelio in the FCC Measuring Broadband America testbed operated by SamKnows and also in the RIPE Atlas testbed. A part of this thesis focuses on characterizing CGNs in Mobile Network Operators (MNOs). We develop a measuring tool, called CGNWatcher that executes a number of active tests to fully characterize CGN deployments in MNOs. The CGNWatcher tool systematically tests more than 30 behavioural requirements of NATs defined by the Internet Engineering Task Force (IETF) and also multiple CGN behavioural metrics. We deploy CGNWatcher in MONROE and performed large measurement campaigns to characterize the real CGN deployments of the MNOs serving the MONROE nodes. We perform a large measurement campaign using the tools described above, recruiting over 6,000 users, from 65 different countries and over 280 ISPs. We validate our results with the ISPs at the IP level and, reported to the ground truth we collected. To the best of our knowledge, this represents the largest active measurement study of (confirmed) NAT or CGN deployments at the IP level in fixed and mobile networks to date. As part of the thesis, we characterize roaming across Europe. The goal of the experiment was to try to understand if the MNO changes CGN while roaming, for this reason, we run a series of measurements that enable us to identify the roaming setup, infer the network configuration for the 16 MNOs that we measure and quantify the end-user performance for the roaming configurations which we detect. We build a unique roaming measurement platform deployed in six countries across Europe. Using this platform, we measure different aspects of international roaming in 3G and 4G networks, including mobile network configuration, performance characteristics, and content discrimination. We find that operators adopt common approaches to implementing roaming, resulting in additional latency penalties of 60 ms or more, depending on geographical distance. Considering content accessibility, roaming poses additional constraints that leads to only minimal deviations when accessing content in the original country. However, geographical restrictions in the visited country make the picture more complicated and less intuitive. Results included in this thesis would provide useful input for application, protocol designers, ISPs and researchers that aim to make their applications and protocols to work across the modern Internet. Programa de Doctorado en Ingeniería Telemática por la Universidad Carlos III de Madrid Presidente: Gonzalo Camarillo González.- Secretario: María Carmen Guerrero López.- Vocal: Andrés García Saavedra

Mención Internacional en el título de doctor The proliferation of handheld devices and the pressing need of location-based services call for precise and accurate ubiquitous geographic mobile positioning that can serve a vast set of devices. Despite the large investments and efforts in academic and industrial communities, a pin-point solution is however still far from reality. Mobile devices mainly rely on Global Navigation Satellite System (GNSS) to position themselves. GNSS systems are known to perform poorly in dense urban areas and indoor environments, where the visibility of GNSS satellites is reduced drastically. In order to ensure interoperability between the technologies used indoor and outdoor, a pervasive positioning system should still rely on GNSS, yet complemented with technologies that can guarantee reliable radio signals in indoor scenarios. The key fact that we exploit is that GNSS signals are made of data with timing information. We then investigate solutions where opportunistic timing signals can be extracted out of terrestrial technologies. These signals can then be used as additional inputs of the multi-lateration problem. Thus, we design and investigate a hybrid system that combines range measurements from the Global Positioning System (GPS), the world’s most utilized GNSS system, and terrestrial technologies; the most suitable one to consider in our investigation is WiFi, thanks to its large deployment in indoor areas. In this context, we first start investigating standalone WiFi Time-of-flight (ToF)-based localization. Time-of-flight echo techniques have been recently suggested for ranging mobile devices overWiFi radios. However, these techniques have yielded only moderate accuracy in indoor environments because WiFi ToF measurements suffer from extensive device-related noise which makes it challenging to differentiate between direct path from non-direct path signal components when estimating the ranges. Existing multipath mitigation techniques tend to fail at identifying the direct path when the device-related Gaussian noise is in the same order of magnitude, or larger than the multipath noise. In order to address this challenge, we propose a new method for filtering ranging measurements that is better suited for the inherent large noise as found in WiFi radios. Our technique combines statistical learning and robust statistics in a single filter. The filter is lightweight in the sense that it does not require specialized hardware, the intervention of the user, or cumbersome on-site manual calibration. This makes the method we propose as the first contribution of the present work particularly suitable for indoor localization in large-scale deployments using existing legacy WiFi infrastructures. We evaluate our technique for indoor mobile tracking scenarios in multipath environments, and, through extensive evaluations across four different testbeds covering areas up to 1000m2, the filter is able to achieve a median ranging error between 1:7 and 2:4 meters. The next step we envisioned towards preparing theoretical and practical basis for the aforementioned hybrid positioning system is a deep inspection and investigation of WiFi and GPS ToF ranges, and initial foundations of single-technology self-localization. Self-localization systems based on the Time-of-Flight of radio signals are highly susceptible to noise and their performance therefore heavily rely on the design and parametrization of robust algorithms. We study the noise sources of GPS and WiFi ToF ranging techniques and compare the performance of different selfpositioning algorithms at a mobile node using those ranges. Our results show that the localization error varies greatly depending on the ranging technology, algorithm selection, and appropriate tuning of the algorithms. We characterize the localization error using real-world measurements and different parameter settings to provide guidance for the design of robust location estimators in realistic settings. These tools and foundations are necessary to tackle the problem of hybrid positioning system providing high localization capabilities across indoor and outdoor environments. In this context, the lack of a single positioning system that is able the fulfill the specific requirements of diverse indoor and outdoor applications settings has led the development of a multitude of localization technologies. Existing mobile devices such as smartphones therefore commonly rely on a multi-RAT (Radio Access Technology) architecture to provide pervasive location information in various environmental contexts as the user is moving. Yet, existing multi-RAT architectures consider the different localization technologies as monolithic entities and choose the final navigation position from the RAT that is foreseen to provide the highest accuracy in the particular context. In contrast, we propose in this work to fuse timing range (Time-of-Flight) measurements of diverse radio technologies in order to circumvent the limitations of the individual radio access technologies and improve the overall localization accuracy in different contexts. We introduce an Extended Kalman filter, modeling the unique noise sources of each ranging technology. As a rich set of multiple ranges can be available across different RATs, the intelligent selection of the subset of ranges with accurate timing information is critical to achieve the best positioning accuracy. We introduce a novel geometrical-statistical approach to best fuse the set of timing ranging measurements. We also address practical problems of the design space, such as removal of WiFi chipset and environmental calibration to make the positioning system as autonomous as possible. Experimental results show that our solution considerably outperforms the use of monolithic technologies and methods based on classical fault detection and identification typically applied in standalone GPS technology. All the contributions and research questions described previously in localization and positioning related topics suppose full knowledge of the anchors positions. In the last part of this work, we study the problem of deriving proximity metrics without any prior knowledge of the positions of the WiFi access points based on WiFi fingerprints, that is, tuples of WiFi Access Points (AP) and respective received signal strength indicator (RSSI) values. Applications that benefit from proximity metrics are movement estimation of a single node over time, WiFi fingerprint matching for localization systems and attacks on privacy. Using a large-scale, real-world WiFi fingerprint data set consisting of 200,000 fingerprints resulting from a large deployment of wearable WiFi sensors, we show that metrics from related work perform poorly on real-world data. We analyze the cause for this poor performance, and show that imperfect observations of APs with commodity WiFi clients in the neighborhood are the root cause. We then propose improved metrics to provide such proximity estimates, without requiring knowledge of location for the observed AP. We address the challenge of imperfect observations of APs in the design of these improved metrics. Our metrics allow to derive a relative distance estimate based on two observed WiFi fingerprints. We demonstrate that their performance is superior to the related work metrics. This work has been supported by IMDEA Networks Institute Programa Oficial de Doctorado en Ingeniería Telemática Presidente: Francisco Barceló Arroyo.- Secretario: Paolo Casari.- Vocal: Marco Fiore

Este trabajo se enmarca dentro del vasto contexto de Ciudades Inteligentes, y se centra en el área de la conducción inteligente de vehículos, tanto en zonas urbanas como interurbanas, mediante la recogida de datos en tiempo real, medidos con sensores, por parte de los propios conductores, así como de datos capturados mediante simulación. El objetivo de este trabajo es doble. Por un lado, el estudio y aplicación de las diferentes técnicas y métodos de detección de valores atípicos en bases de datos multivariantes, además de una comparativa entre ellos mediante las pruebas llevadas a cabo con datos de tráfico real. Y por otro lado, establecer una relación entre las situaciones anómalas de tráfico, como puedan ser atascos o accidentes, con los valores atípicos multivariantes encontrados. La detección de valores atípicos representa una de las tareas más importantes a la hora de realizar cualquier análisis de datos, sea cual sea el dominio o área de estudio, ya que entre sus funciones primordiales se encuentra el descubrir información útil, que resulta de gran valor, y que por lo general queda oculta por la alta dimensión de los datos. Con el uso de mecanismos de detección de valores atípicos junto con métodos de clasificación supervisada, se va a poder llevar a cabo el reconocimiento de elementos de la infraestructura vial urbana como pueden ser rotondas, pasos de cebra, cruces o semáforos. This work is related to the Smart Cities context, and it focuses on the area of intelligent vehicle driving, both in urban and interurban areas, through the collection of real-time sensed data by the drivers themselves, as well as data collected in a simulator. The goal of this paper is twofold. On the one hand, the study and application of the different techniques and methods of outliers detection in multivariate databases, as well as a comparison between them through the tests carried out with real traffic data. And on the other hand, to establish a relation between anomalous traffic situations, such as traffic jams or accidents, with the multivariate outliers found. Outliers detection represents one of the most important tasks when performing any data analysis, regardless of the domain or area of study, since among its fundamental functions is to discover useful and valuable information that usually is hidden by the high dimensionality of the data. By means of using outliers detection mechanisms together with data classification methods, the recognition of elements of urban infrastructure such as roundabouts, zebra crossing or traffic lights will be carried out. Este trabajo ha sido financiado por el Ministerio de Economía, Industria y Competitividad, con la ayuda FPI: BES-2014-070462 Programa Oficial de Doctorado en Ingeniería Telemática Presidente: Natividad Martínez Madrid.- Secretario: Jesús Arias Fisteus.- Vocal: Norberto Fernández García

Mención Internacional en el título de doctor The consumption of multimedia content is moving from a residential environment to mobile phones. Mobile data traffic, driven mostly by video demand, is increasing rapidly and wireless spectrum is becoming a more and more scarce resource. This makes it highly important to operate mobile networks efficiently. To tackle this, recent developments in anticipatory networking schemes make it possible to to predict the future capacity of mobile devices and optimize the allocation of the limited wireless resources. Further, optimizing Quality of Experience—smooth, quick, and high quality playback—is more difficult in the mobile setting, due to the highly dynamic nature of wireless links. A key requirement for achieving, both anticipatory networking schemes and QoE optimization, is estimating the available bandwidth of mobile devices. Ideally, this should be done quickly and with low overhead. In summary, we propose a series of improvements to the delivery of multimedia over mobile networks. We do so, be identifying inefficiencies in the interconnection of mobile operators with the servers hosting content, propose an algorithm to opportunistically create frequent capacity estimations suitable for use in resource optimization solutions and finally propose another algorithm able to estimate the bandwidth class of a device based on minimal traffic in order to identify the ideal streaming quality its connection may support before commencing playback. The main body of this thesis proposes two lightweight algorithms designed to provide bandwidth estimations under the high constraints of the mobile environment, such as and most notably the usually very limited traffic quota. To do so, we begin with providing a thorough overview of the communication path between a content server and a mobile device. We continue with analysing how accurate smartphone measurements can be and also go in depth identifying the various artifacts adding noise to the fidelity of on device measurements. Then, we first propose a novel lightweight measurement technique that can be used as a basis for advanced resource optimization algorithms to be run on mobile phones. Our main idea leverages an original packet dispersion based technique to estimate per user capacity. This allows passive measurements by just sampling the existing mobile traffic. Our technique is able to efficiently filter outliers introduced by mobile network schedulers and phone hardware. In order to asses and verify our measurement technique, we apply it to a diverse dataset generated by both extensive simulations and a week-long measurement campaign spanning two cities in two countries, different radio technologies, and covering all times of the day. The results demonstrate that our technique is effective even if it is provided only with a small fraction of the exchanged packets of a flow. The only requirement for the input data is that it should consist of a few consecutive packets that are gathered periodically. This makes the measurement algorithm a good candidate for inclusion in OS libraries to allow for advanced resource optimization and application-level traffic scheduling, based on current and predicted future user capacity. We proceed with another algorithm that takes advantage of the traffic generated by short-lived TCP connections, which form the majority of the mobile connections, to passively estimate the currently available bandwidth class. Our algorithm is able to extract useful information even if the TCP connection never exits the slow start phase. To the best of our knowledge, no other solution can operate with such constrained input. Our estimation method is able to achieve good precision despite artifacts introduced by the slow start behavior of TCP, mobile scheduler and phone hardware. We evaluate our solution against traces collected in 4 European countries. Furthermore, the small footprint of our algorithm allows its deployment on resource limited devices. Finally, in an attempt to face the rapid traffic increase, mobile application developers outsource their cloud infrastructure deployment and content delivery to cloud computing services and content delivery networks. Studying how these services, which we collectively denote Cloud Service Providers (CSPs), perform over Mobile Network Operators (MNOs) is crucial to understanding some of the performance limitations of today’s mobile apps. To that end, we perform the first empirical study of the complex dynamics between applications, MNOs and CSPs. First, we use real mobile app traffic traces that we gathered through a global crowdsourcing campaign to identify the most prevalent CSPs supporting today’s mobile Internet. Then, we investigate how well these services interconnect with major European MNOs at a topological level, and measure their performance over European MNO networks through a month-long measurement campaign on the MONROE mobile broadband testbed. We discover that the top 6 most prevalent CSPs are used by 85% of apps, and observe significant differences in their performance across different MNOs due to the nature of their services, peering relationships with MNOs, and deployment strategies. We also find that CSP performance in MNOs is affected by inflated path length, roaming, and presence of middleboxes, but not influenced by the choice of DNS resolver. We also observe that the choice of operator’s Point of Presence (PoP) may inflate by at least 20% the delay towards popular websites. This work has been supported by IMDEA Networks Institute. Programa Oficial de Doctorado en Ingeniería Telemática Presidente: Ahmed Elmokashfi.- Secretario: Rubén Cuevas Rumín.- Vocal: Paolo Dini

Mención Internacional en el título de doctor Mobile user devices’ market has experi-enced an exponential growth worldwide over the last decade, and wireless communications are the main driver for the next generation of 5G networks. The ubiquity of battery-powered connected devices makes energy efficiency a major research issue. While most studies assumed that network interfaces dominate the energy consumption of wireless communications, a recent work unveils that the frame processing carried out by the device could drain as much energy as the interface itself for many devices. This discovery poses doubts on prior energy models for wireless communications and forces us to reconsider existing energy-saving schemes. From this standpoint, this thesis is de-voted to the study of the energy efficiency of mobile user devices at multiple layers. To that end, we assemble a comprehensive en-ergy measurement framework, and a robust methodology, to be able to characterise a wide range of mobile devices, as well as individual parts of such devices. Building on this, we first delve into the en-ergy consumption of frame processing within the devices’ protocol stack. Our results identify the CPU as the leading cause of this energy consumption. Moreover, we discover that the characterisation of the energy toll ascribed to the device is much more complex than the previous work showed. Devices with complex CPUs (several frequencies and sleep states) require novel methodologies and models to successfully characterise their consumption. We then turn our attention to lower levels of the communication stack by investigating the behaviour of idle WiFi interfaces. Due to the design of the 802.11 protocol, together with the growing trend of network densification, WiFi devices spend a long time receiving frames addressed to other devices when they might be dormant. In order to mitigate this issue, we study the timing constraints of a commercial WiFi card, which is developed into a standard-compliant algorithm that saves energy during such transmissions. At a higher level, rate adaptation and power control techniques adapt data rate and output power to the channel conditions. However, these have been typically studied with other metrics rather than energy efficiency in mind (i.e., performance figures such as throughput and capacity). In fact, our analyses and sim-ulations unveil an inherent trade-off between throughput and energy efficiency maximisa-tion in 802.11. We show that rate adaptation and power control techniques may incur inef-ficiencies at mode transitions, and we provide energy-aware heuristics to make such decisions following a conservative approach. Finally, our research experience on simula-tion methods pointed us towards the need for new simulation tools commited to the middle-way approach: less specificity than complex network simulators in exchange for easier and faster prototyping. As a result, we developed a process-oriented and trajectory-based discrete-event simulation package for the R language, which is designed as a easy-to-use yet pow-erful framework with automatic monitoring capabilities. The use of this simulator in net-working is demonstrated through the energy modelling of an Internet-of-Things scenario with thousands of metering devices in just a few lines of code. El mercado de los dispositivos de usuario móviles ha experimentado un crecimiento exponencial a nivel mundial en la última década, y las comunicaciones inalámbricas son el principal motor de la siguiente generación de redes 5G. La ubicuidad de estos dispos-itivos alimentados por baterías hace de la eficiencia energética un importante tema de investigación. Mientras muchos estudios asumían que la interfaz de red domina el consumo energético de las comuni-caciones inalámbricas, un trabajo reciente revela que el procesado de tramas que se lleva a cabo en el disposi-tivo podría gastar tanta energía como la propia interfaz para muchos dispositivos. Este descubrimiento plantea dudas sobre los anteriores modelos energéticos para comunicaciones inalámbricas y nos obliga a reconsid-erar los esquemas de ahorro energético existentes. Desde este punto de vista, esta tesis está dedicada al estudio de la eficiencia energética de dispositivos de usuario móviles en múltiples capas. Para ello, se construye un completo sistema de medida de energía, y una metodología robusta, capaz de caracterizar un amplio rango de dispositivos móviles, así como partes individuales de tales dispositivos. A partir de esto, en primer lugar se profundiza en el consumo energético del procesamiento de tramas en la pila de protocolos de los dispositivos. Nuestros resul-tados identifican a la CPU como principal causa de tal consumo. Además, se descubre que la caracterización de la cuota energética adscrita al dispositivo es mucho más compleja que lo mostrado por el trabajo ante-rior. Los dispositivos con CPU complejas (múltiples frecuencias y modos de apagado) requieren nuevas metodologías y modelos para caracterizar su consumo de manera existosa. En este punto, volvemos nuestra atención hacia niveles más bajos de la pila de comunicaciones para investigar el comportamiento de las interfaces WiFi en estado inactivo. Debido al diseño del protocolo 802.11, junto con la tendencia creciente hacia la densifi-cación de las redes, los dispositivos WiFi pasan mucho tiempo recibiendo tramas destinadas a otros dispos-itivos cuando podrían estar apagados. Para mitigar este problema, se estudian las limitaciones temporales de una tarjeta WiFi comercial, lo que posteriormente se utiliza para desarrollar un algoritmo conforme con el estándar que es capaz de ahorrar energía durante dichas transmisiones. A un nivel más alto, las técnicas de adaptación de tasa y control de potencia adaptan la tasa de datos y la potencia de salida a las condiciones del canal. No obstante, estas técnicas han sido típicamente es-tudiadas con otras métricas en mente (i.e., figuras de rendimiento como la tasa total y la capacidad). De hecho, nuestros análisis y simulaciones desvelan un conflicto entre la maximización de la tasa total y la efi-ciencia energética en 802.11. Se muestra que las técni-cas de adaptación de tasa y control de potencia pueden incurrir en ineficiencias en los cambios de modo, y se proporcionan heurísticos para tomar tales decisiones de un modo conservador y eficiente energéticamente. Finalmente, nuestra experiencia investigadora en métodos de simulación nos hizo conscientes de la necesidad de nuevas herramientas de simulación comprometidas con un enfoque intermedio: menos especificidad que los complejos simuladores de re-des a cambio de facilidad y rapidez en el prototipado. Como resultado, se desarrolló un paquete de simu-lación por eventos discretos para el lenguaje R orien-tado a procesos y basado en trayectorias, el cual está diseñado como una herramienta fácil de utilizar a la par que potente con capacidad de monitorización au-tomática integrada. El uso de este simulador en redes se demuestra mediante el modelado en energía de un escenario de la Internet de las Cosas con miles de dis-positivos de medida en tan solo unas pocas líneas de código. Programa Oficial de Doctorado en Ingeniería Telemática Presidente: Juan Manuel López Soler.- Secretario: Francisco Valera Pintor.- Vocal: Paul Horatiu Patras

Mención Internacional en el título de doctor While connecting end-users worldwide, the Internet increasingly promotes local development by making challenges much simpler to overcome, regardless of the field in which it is used: governance, economy, education, health, etc. However, African Network Information Centre (AfriNIC), the Regional Internet Registry (RIR) of Africa, is characterized by the lowest Internet penetration: 28.6% as of March 2017 compared to an average of 49.7% worldwide according to the International Telecommunication Union (ITU) estimates [139]. Moreover, end-users experience a poor Quality of Service (QoS) provided at high costs. It is thus of interest to enlarge the Internet footprint in such under-connected regions and determine where the situation can be improved. Along these lines, this doctoral thesis thoroughly inspects, using both active and passive data analysis, the critical aspects of the African Internet ecosystem and outlines the milestones of a methodology that could be adopted for achieving similar purposes in other developing regions. The thesis first presents our efforts to help build measurements infrastructures for alleviating the shortage of a diversified range of Vantage Points (VPs) in the region, as we cannot improve what we can not measure. It then unveils our timely and longitudinal inspection of the African interdomain routing using the enhanced RIPE Atlas measurements infrastructure for filling the lack of knowledge of both IPv4 and IPv6 topologies interconnecting local Internet Service Providers (ISPs). It notably proposes reproducible data analysis techniques suitable for the treatment of any set of similar measurements to infer the behavior of ISPs in the region. The results show a large variety of transit habits, which depend on socio-economic factors such as the language, the currency area, or the geographic location of the country in which the ISP operates. They indicate the prevailing dominance of ISPs based outside Africa for the provision of intracontinental paths, but also shed light on the efforts of stakeholders for traffic localization. Next, the thesis investigates the causes and impacts of congestion in the African IXP substrate, as the prevalence of this endemic phenomenon in local Internet markets may hinder their growth. Towards this end, Ark monitors were deployed at six strategically selected local Internet eXchange Points (IXPs) and used for collecting Time-Sequence Latency Probes (TSLP) measurements during a whole year. The analysis of these datasets reveals no evidence of widespread congestion: only 2.2% of the monitored links experienced noticeable indication of congestion, thus promoting peering. The causes of these events were identified during IXP operator interviews, showing how essential collaboration with stakeholders is to understanding the causes of performance degradations. As part of the Internet Society (ISOC) strategy to allow the Internet community to profile the IXPs of a particular region and monitor their evolution, a route-collector data analyzer was then developed and afterward, it was deployed and tested in AfriNIC. This open source web platform titled the “African” Route-collectors Data Analyzer (ARDA) provides metrics, which picture in real-time the status of interconnection at different levels, using public routing information available at local route-collectors with a peering viewpoint of the Internet. The results highlight that a small proportion of Autonomous System Numbers (ASNs) assigned by AfriNIC (17 %) are peering in the region, a fraction that remained static from April to September 2017 despite the significant growth of IXPs in some countries. They show how ARDA can help detect the impact of a policy on the IXP substrate and help ISPs worldwide identify new interconnection opportunities in Africa, the targeted region. Since broadening the underlying network is not useful without appropriately provisioned services to exploit it, the thesis then delves into the availability and utilization of the web infrastructure serving the continent. Towards this end, a comprehensive measurement methodology is applied to collect data from various sources. A focus on Google reveals that its content infrastructure in Africa is, indeed, expanding; nevertheless, much of its web content is still served from the United States (US) and Europe, although being the most popular content source in many African countries. Further, the same analysis is repeated across top global and regional websites, showing that even top African websites prefer to host their content abroad. Following that, the primary bottlenecks faced by Content Providers (CPs) in the region such as the lack of peering between the networks hosting our probes and poorly configured DNS resolvers are explored to outline proposals for further ISP and CP deployments. Considering the above, an option to enrich connectivity and incentivize CPs to establish a presence in the region is to interconnect ISPs present at isolated IXPs by creating a distributed IXP layout spanning the continent. In this respect, the thesis finally provides a four-step interconnection scheme, which parameterizes socio-economic, geographical, and political factors using public datasets. It demonstrates that this constrained solution doubles the percentage of continental intra-African paths, reduces their length, and drastically decreases the median of their Round Trip Times (RTTs) as well as RTTs to ASes hosting the top 10 global and top 10 regional Alexa websites. We hope that quantitatively demonstrating the benefits of this framework will incentivize ISPs to intensify peering and CPs to increase their presence, for enabling fast, affordable, and available access at the Internet frontier. Programa Oficial de Doctorado en Ingeniería Telemática Presidente: David Fernández Cambronero.- Secretario: Alberto García Martínez.- Vocal: Cristel Pelsser

Mención Internacional en el título de doctor Cooperative short-range communication schemes provide powerful tools to solve interference and resource shortage problems in wireless access networks. With such schemes, a mobile node with excellent cellular connectivity can momentarily accept to relay traffic for its neighbors experiencing poor radio conditions and use Device-to-Device (D2D) communications to accomplish the task. This thesis provides a novel and comprehensive analytical framework that allows evaluating the effects of D2D communications in access networks in terms of spectrum and energy efficiency. The analysis covers the cases in which D2D communications use the same bandwidth of legacy cellular users (in-band D2D) or a different one (out-band D2D) and leverages on the characterization of underlying queueing systems and protocols to capture the complex intertwining of short-range and legacy WiFi and cellular communications. The analysis also unveils how D2D affects the use and scope of other optimization techniques used for, e.g., interference coordination and fairness in resource distribution. Indeed, characterizing the performance of D2D-enabled wireless access networks plays an essential role in the optimization of system operation and, as a consequence, permits to assess the general applicability of D2D solutions. With such characterization, we were able to design several mechanisms that improve system capabilities. Specifically, we propose bandwidth resource management techniques for controlling interference when cellular users and D2D pairs share the same spectrum, we design advanced and energy-aware access selection mechanisms, we show how to adopt D2D communications in conjunction with interference coordination schemes to achieve high and fair throughputs, and we discuss on end-to-end fairness—beyond the use of access network resources—when D2D communications is adopted in C-RAN. The results reported in this thesis show that identifying performance bottlenecks is key to properly control network operation, and, interestingly, bottlenecks may not be represented just by wireless resources when end-to-end fairness is of concern. Programa Oficial de Doctorado en Ingeniería Telemática Presidente: Marco Ajmone Marsan.- Secretario: Miquel Payaró Llisterri.- Vocal: Omer Gurewitz

Mención Internacional en el título de doctor Mobile cellular networks are complex system whose behavior is characterized by the superposition of several random phenomena, most of which, related to human activities, such as mobility, communications and network usage. However, when observed in their totality, the many individual components merge into more deterministic patterns and trends start to be identifiable and predictable. In this thesis we analyze a recent branch of network optimization that is commonly referred to as anticipatory networking and that entails the combination of prediction solutions and network optimization schemes. The main intuition behind anticipatory networking is that knowing in advance what is going on in the network can help understanding potentially severe problems and mitigate their impact by applying solution when they are still in their initial states. Conversely, network forecast might also indicate a future improvement in the overall network condition (i.e. load reduction or better signal quality reported from users). In such a case, resources can be assigned more sparingly requiring users to rely on buffered information while waiting for the better condition when it will be more convenient to grant more resources. In the beginning of this thesis we will survey the current anticipatory networking panorama and the many prediction and optimization solutions proposed so far. In the main body of the work, we will propose our novel solutions to the problem, the tools and methodologies we designed to evaluate them and to perform a real world evaluation of our schemes. By the end of this work it will be clear that not only is anticipatory networking a very promising theoretical framework, but also that it is feasible and it can deliver substantial benefit to current and next generation mobile networks. In fact, with both our theoretical and practical results we show evidences that more than one third of the resources can be saved and even larger gain can be achieved for data rate enhancements. Programa Oficial de Doctorado en Ingeniería Telemática Presidente: Albert Banchs Roca.- Presidente: Pablo Serrano Yañez-Mingot.- Secretario: Jorge Ortín Gracia.- Vocal: Guevara Noubir