Search

Traditional machine learning systems were designed in a centralized manner. In such designs, the central entity maintains both the machine learning model and the data used to adjust the model's parameters. As data centralization yields privacy issues, Federated Learning was introduced to reduce data sharing and have a central server coordinate the learning of multiple devices. While Federated Learning is more decentralized, it still relies on a central entity that may fail or be subject to attacks, provoking the failure of the whole system. Then, Decentralized Federated Learning removes the need for a central server entirely, letting participating processes handle the coordination of the model construction. This distributed control urges studying the possibility of malicious attacks by the participants themselves. While poisoning attacks on Federated Learning have been extensively studied, their effects in Decentralized Federated Learning did not get the same level of attention. Our work is the first to propose a methodology to assess poisoning attacks in Decentralized Federated Learning in both churn free and churn prone scenarios. Furthermore, in order to evaluate our methodology on a case study representative for gossip learning we extended the gossipy simulator with an attack injector module.

A number of projects in computer science around the world have contributed to build federated experimental facilities providing access to a large set of distributed compute, storage and network resources for the research community. Several tools have been developed to provide users an easy access to the federated testbeds. This paper presents the architecture of the new version of the MySlice web portal, that has evolved from a prototype to a production ready software.

Network Functions Virtualization (NFV) aims to support service providers to deploy various services in a more agile and cost-effective way. However, the softwarization and cloudification of network functions can result in severe congestion and low network performance. In this paper, we propose a solution to address this issue. We analyze and solve the online load balancing problem using multipath routing in NFV to optimize network performance in response to the dynamic changes of user demands. In particular, we first formulate the optimization problem of load balancing as a mixed integer linear program for achieving the optimal solution. We then develop the ORBIT algorithm that solves the online load balancing problem. The performance guarantee of ORBIT is analytically proved in comparison with the optimal offline solution. The experiment results on real-world datasets show that ORBIT performs very well for distributing traffic of each service demand across multipaths without knowledge of future demands, especially under high-load conditions.

This paper proposes Navigo, a location based packet forwarding mechanism for vehicular Named Data Networking (NDN). Navigo takes a radically new approach to address the challenges of frequent connectivity disruptions and sudden network changes in a vehicle network. Instead of forwarding packets to a specific moving car, Navigo aims to fetch specific pieces of data from multiple potential carriers of the data. The design provides (1) a mechanism to bind NDN data names to the producers' geographic area(s); (2) an algorithm to guide Interests towards data producers using a specialized shortest path over the road topology; and (3) an adaptive discovery and selection mechanism that can identify the best data source across multiple geographic areas, as well as quickly react to changes in the V2X network.

Intense vehicular traffic is recognized as a global societal problem, with a multifaceted influence on the quality of life of a person. Intelligent Transportation Systems (ITS) can play an important role in combating such problem, decreasing pollution levels and, consequently, their negative effects. One of the goals of ITSs, in fact, is that of controlling traffic flows, measuring traffic states, providing vehicles with routes that globally pursue low pollution conditions. How such systems measure and enforce given traffic states has been at the center of multiple research efforts in the past few years. Although many different solutions have been proposed, very limited effort has been devoted to exploring the potential of social network analysis in such context. Social networks, in general, provide direct feedback from people and, as such, potentially very valuable information. A post that tells, for example, how a person feels about pollution at a given time in a given location, could be put to good use by an environment aware ITS aiming at minimizing contaminant emissions in residential areas. This work verifies the feasibility of using pollution related social network feeds into ITS operations. In particular, it concentrates on understanding how reliable such information is, producing an analysis that confronts over 1,500,000 posts and pollution data obtained from on-the- field sensors over a one-year span., Comment: 10 pages, 15 figures, Transaction Format

We consider a simple two-player game involving a large incumbent and small entrant into a cellular wireless access provider marketplace. The entrant's customers must pay roaming charges. We assume that the roaming charges are regulated, because if they are dictated by the incumbent then they could be set so high so as to be a barrier to entry in the marketplace. The game is studied at its Nash equilibrium. A roaming charge is identified that is arguably fair in the sense that revenues for the access providers are proportionate to their infrastructure costs.

Despite its existing incentives for leecher cooperation, BitTorrent file sharing fundamentally relies on the presence of seeder peers. Seeder peers essentially operate outside the BitTorrent incentives, with two caveats: slow downlinks lead to increased numbers of "temporary" seeders (who left their console, but will terminate their seeder role when they return), and the copyright liability boon that file segmentation offers for permanent seeders. Using a simple epidemic model for a two-segment BitTorrent swarm, we focus on the BitTorrent rule to disseminate the (locally) rarest segments first. With our model, we show that the rarest-segment first rule minimizes transition time to seeder (complete file acquisition) and equalizes the segment populations in steady-state. We discuss how alternative dissemination rules may {\em beneficially increase} file acquisition times causing leechers to remain in the system longer (particularly as temporary seeders). The result is that leechers are further enticed to cooperate. This eliminates the threat of extinction of rare segments which is prevented by the needed presence of permanent seeders. Our model allows us to study the corresponding trade-offs between performance improvement, load on permanent seeders, and content availability, which we leave for future work. Finally, interpreting the two-segment model as one involving a rare segment and a "lumped" segment representing the rest, we study a model that jointly considers control of rare segments and different uplinks causing "choking," where high-uplink peers will not engage in certain transactions with low-uplink peers., Comment: 18 pages, 6 figures, A shorter version of this paper that did not include the N-segment lumped model was presented in May 2011 at IEEE ICC, Kyoto

In this paper, we consider medium access control of local area networks (LANs) under limited-information conditions as befits a distributed system. Rather than assuming "by rule" conformance to a protocol designed to regulate packet-flow rates (e.g., CSMA windowing), we begin with a non-cooperative game framework and build a dynamic altruism term into the net utility. The effects of altruism are analyzed at Nash equilibrium for both the ALOHA and CSMA frameworks in the quasistationary (fictitious play) regime. We consider either power or throughput based costs of networking, and the cases of identical or heterogeneous (independent) users/players. In a numerical study we consider diverse players, and we see that the effects of altruism for similar players can be beneficial in the presence of significant congestion, but excessive altruism may lead to underuse of the channel when demand is low.

We consider wireless mesh networks and the problem of routing end-to-end traffic over multiple paths for the same origin-destination pair with minimal interference. We introduce a heuristic for path determination with two distinguishing characteristics. First, it works by refining an extant set of paths, determined previously by a single- or multi-path routing algorithm. Second, it is totally local, in the sense that it can be run by each of the origins on information that is available no farther than the node's immediate neighborhood. We have conducted extensive computational experiments with the new heuristic, using AODV and OLSR, as well as their multi-path variants, as underlying routing methods. For two different CSMA settings (as implemented by 802.11) and one TDMA setting running a path-oriented link scheduling algorithm, we have demonstrated that the new heuristic is capable of improving the average throughput network-wide. When working from the paths generated by the multi-path routing algorithms, the heuristic is also capable to provide a more evenly distributed traffic pattern.

We consider wireless mesh networks and the problem of scheduling the links of a given set of routes under the assumption of a heavy-traffic pattern. We assume some TDMA protocol provides a background of synchronized time slots and seek to schedule the routes' links to maximize the number of packets that get delivered to their destinations per time slot. Our approach is to construct an undirected graph G and to heuristically obtain node multicolorings for G that can be turned into efficient link schedules. In G each node represents a link to be scheduled and the edges are set up to represent every possible interference for any given set of interference assumptions. We present two multicoloring-based heuristics and study their performance through extensive simulations. One of the two heuristics is based on relaxing the notion of a node multicoloring by dynamically exploiting the availability of communication opportunities that would otherwise be wasted. We have found that, as a consequence, its performance is significantly superior to the other's.

This paper introduces a new channel selection strategy for reliable contentionaware data dissemination in multi-hop cognitive radio network. The key challenge here is to select channels providing a good tradeoff between connectivity and contention. In other words, channels with good opportunities for communication due to (1) low primary radio nodes (PRs) activities, and (2) limited contention of cognitive ratio nodes (CRs) acceding that channel, have to be selected. Thus, by dynamically exploring residual resources on channels and by monitoring the number of CRs on a particular channel, SURF allows building a connected network with limited contention where reliable communication can take place. Through simulations, we study the performance of SURF when compared with three other related approaches. Simulation results confirm that our approach is effective in selecting the best channels for efficient and reliable multi-hop data dissemination.

In this paper, we propose a cognitive radio based Internet access framework for disaster response network deployment in challenged environments. The proposed architectural framework is designed to help the existent but partially damaged networks to restore their connectivity and to connect them to the global Internet. This architectural framework provides the basis to develop algorithms and protocols for the future cognitive radio network deployments in challenged environments.

In this paper, we propose an adaptive and occupancy-based channel selection for unreliable cognitive radio networks.

Assessing mobility in a thorough fashion is a crucial step toward more efficient mobile network design. Recent research on mobility has focused on two main points: analyzing models and studying their impact on data transport. These works investigate the consequences of mobility. In this paper, instead, we focus on the causes of mobility. Starting from established research in sociology, we propose SIMPS, a mobility model of human crowd motion. This model defines two complimentary behaviors, namely socialize and isolate, that regulate an individual with regard to her/his own sociability level. SIMPS leads to results that agree with scaling laws observed both in small-scale and large-scale human motion. Although our model defines only two simple individual behaviors, we observe many emerging collective behaviors (group formation/splitting, path formation, and evolution). To our knowledge, SIMPS is the first model in the networking community that tackles the roots governing mobility.

This paper introduces Prawn, a tool for prototyping communication protocols over IEEE 802.11 networks. Prawn allows researchers to conduct both functional assessment and performance evaluation as an inherent part of the protocol design process. Since Prawn runs on real IEEE 802.11 nodes, prototypes can be evaluated and adjusted under realistic conditions. Once the prototype has been extensively tested and thoroughly validated, and its functional design tuned accordingly, it is then ready for implementation. Prawn facilitates prototype development by providing: (i) a set of building blocks that implement common functions needed by a wide range of wireless protocols (e.g., neighbor discovery, link quality assessment, message transmission and reception), and (ii) an API that allows protocol designers to access Prawn primitives. We show through a number of case studies how Prawn supports prototyping as part of protocol design and, as a result of enabling deployment and testing under real-world scenarios, how Prawn provides useful feedback on protocol operation and performance., Comment: Changed content. 12 pages

In the widely used 802.11 standard, the so called performance anomaly is a well known issue. Several works have tried to solve this problem by introducing mechanisms such as packet fragmentation, backoff adaptation, or packet aggregation during a fixed time interval. In this paper, we propose a novel approach solving the performance anomaly problem by packet aggregation using a dynamic time interval, which depends on the busy time of the wireless medium. Our solution differs from other proposition in the literature because of this dynamic time interval, which allows increasing fairness, reactivity, and in some cases efficiency. In this article, we emphasize the performance evaluation of our proposal.