Search

The dynamic and evolving nature of mobile networks necessitates a proactive approach to security, one that goes beyond traditional methods and embraces innovative strategies such as anomaly detection and prediction. This study delves into the realm of mobile network security and reliability enhancement through the lens of anomaly detection and prediction, leveraging K-means clustering on call detail records (CDRs). By analyzing CDRs, which encapsulate comprehensive information about call activities, messaging, and data usage, this research aimed to unveil hidden patterns indicative of anomalous behavior within mobile networks and security breaches. We utilized 14 million one-year CDR records. The mobile network used had deployed the latest network generation, 5G, with various sources of network elements. Through a systematic analysis of historical CDR data, this study offers insights into the underlying trends and anomalies prevalent in mobile network traffic. Furthermore, by harnessing the predictive capabilities of the K-means algorithm, the proposed framework facilitates the anticipation of future anomalies based on learned patterns, thereby enhancing proactive security measures. The findings of this research can contribute to the advancement of mobile network security by providing a deeper understanding of anomalous behavior and effective prediction mechanisms. The utilization of K-means clustering on CDR data offers a scalable and efficient approach to anomaly detection, with 96% accuracy, making it well suited for network reliability and security applications in large-scale mobile networks for 5G networks and beyond. [ABSTRACT FROM AUTHOR]

In recent decades, with the attracting features of mobiles including 4G and 5G, world is getting more connected to mobile communications. This results in the accumulation of large amount of data in the mobile network. The analysis of the network data is very complex but is essential in terms of resource and cost management. The network data analytics include detection of unusual network behaviour due to traffic created by the mobile users and Short Message Service (SMS) spammers. Research to an approach with the same impulsion is creating a new interest in the field of mobile network data analytics using machine learning tools. To attain this, Call Detail Record (CDR) provided by the telecom network industry is utilized. The timely analysis of CDR helps to understand the behaviour of the network due to various activities of mobile users. To analyse CDR, it has to be pre-processed to convert it from the raw data into machine understandable form. The proposed method is mean-normalization pre-processing which is suitable in understanding the behaviour of mobile users’ individual activities like incoming-outgoing calls, incoming-outgoing SMS and internet activity. Later, machine learning tools can be applied to analyse and predict the network anomalies like network traffic and Short.

One major source of uncertainty in accurately estimating human exposure to air pollution is that human subjects move spatiotemporally, and such mobility is usually not considered in exposure estimation. How such mobility impacts exposure estimates at the population and individual level, particularly for subjects with different levels of mobility, remains under-investigated. In addition, a wide range of methods have been used in the past to develop air pollutant concentration fields for related health studies. How the choices of methods impact results of exposure estimation, especially when detailed mobility information is considered, is still largely unknown. In this study, by using a publicly available large cell phone location dataset containing over 35 million location records collected from 310,989 subjects, we investigated the impact of individual subjects’ mobility on their estimated exposures for five chosen ambient pollutants (CO, NO2, SO2, O3 and PM2.5). We also estimated exposures separately for 10 groups of subjects with different levels of mobility to explore how increased mobility impacted their exposure estimates. Further, we applied and compared two methods to develop concentration fields for exposure estimation, including one based on Community Multiscale Air Quality (CMAQ) model outputs, and the other based on the interpolated observed pollutant concentrations using the inverse distance weighting (IDW) method. Our results suggest that detailed mobility information does not have a significant influence on mean population exposure estimate in our sample population, although impacts can be substantial at the individual level. Additionally, exposure classification error due to the use of home-location data increased for subjects that exhibited higher levels of mobility. Omitting mobility could result in underestimation of exposures to traffic-related pollutants particularly during afternoon rush-hour, and overestimate exposures to ozone especially during mid-afternoon. Between CMAQ and IDW, we found that the IDW method generates smooth concentration fields that were not suitable for exposure estimation with detailed mobility data. Therefore, the method for developing air pollution concentration fields when detailed mobility data were to be applied should be chosen carefully. Our findings have important implications for future air pollution health studies.

5G is anticipated to embed an artificial intelligence (AI)-empowerment to adroitly plan, optimize and manage the highly complex network by leveraging data generated at different positions of the network architecture. Outages and situation leading to congestion in a cell pose severe hazard for the network. High false alarms and inadequate accuracy are the major limitations of modern approaches for the anomaly-outage and sudden hype in traffic activity that may result in congestion-detection in mobile cellular networks. This indicates wasting limited resources that ultimately leads to an elevated operational expenditure (OPEX) and also interrupting quality of service (QoS) and quality of experience (QoE). Motivated by the outstanding success of deep learning (DL) technology, our study applies it for detection of the above-mentioned anomalies and also supports mobile edge computing (MEC) paradigm in which core network (CN)'s computations are divided across the cellular infrastructure among different MEC servers (co-located with base stations), to relief the CN. Each server monitors user activities of multiple cells and utilizes L -layer feedforward deep neural network (DNN) fueled by real call detail record (CDR) dataset for anomaly detection. Our framework achieved 98.8% accuracy with 0.44% false positive rate (FPR)-notable improvements that surmount the deficiencies of the old studies. The numerical results explicate the usefulness and dominance of our proposed detector.

Escalating cell outages and congestion—treated as anomalies—cost a substantial revenue loss to the cellular operators and severely affect subscriber quality of experience. State-of-the-art literature applies feed-forward deep neural network at core network (CN) for the detection of above problems in a single cell; however, the solution is impractical as it will overload the CN that monitors thousands of cells at a time. Inspired from mobile edge computing and breakthroughs of deep convolutional neural networks (CNNs) in computer vision research, in this article we split the network into several 100-cell regions each monitored by an edge server; and propose a framework that preprocesses raw call detail records having user activities to create an image-like volume, fed to a CNN model. The framework outputs a multilabeled vector identifying anomalous cell(s). Our results suggest that our solution can detect anomalies with up to 96% accuracy, and is scalable and expandable for industrial Internet of Things environment. [ABSTRACT FROM AUTHOR]

Air pollution epidemiologic and health impact studies often rely on home addresses to estimate individual subject's pollution exposure. In this study, we used detailed cell phone location data, the call detail record (CDR), to account for the impact of spatiotemporal subject mobility on estimates of ambient air pollutant exposure. This approach was applied on a sample with 9886 unique simcard IDs in Shenzhen, China, on one mid-week day in October 2013. Hourly ambient concentrations of six chosen pollutants were simulated by the Community Multi-scale Air Quality model fused with observational data, and matched with detailed location data for these IDs. The results were compared with exposure estimates using home addresses to assess potential exposure misclassification errors. We found the misclassifications errors are likely to be substantial when home location alone is applied. The CDR based approach indicates that the home based approach tends to over-estimate exposures for subjects with higher exposure levels and under-estimate exposures for those with lower exposure levels. Our results show that the cell phone location based approach can be used to assess exposure misclassification error and has the potential for improving exposure estimates in air pollution epidemiology studies. [ABSTRACT FROM AUTHOR]

Human mobility patterns created from mobile phone call detail records (CDRs) can provide an essential resource in data-poor environments to monitor the effects of health outbreaks. Analysis of this data can be instrumental for understanding the movement pattern of populations allowing governments to set and refine policies to respond to community health risks. Building on CDR mobility analysis techniques, this research set out to test whether combining CDR mobility indicators with socio-economic information can illustrate differences between different socio-economic groups' exposure risks to COVID-19. The work focuses on the Western Area of Sierra Leone which houses the capital Freetown because it lacks existing mobility data and therefore can be a great example of how CDR can be transformed for this use. To determine mobility patterns, we applied the radius of gyration, regularity of movement, and motif types analytics commonly used in CDR research. We then applied a clustering algorithm to these results to understand user trends. Then we compared the results of the three methods with socio-economic status determined from census data in the same geography. The results show the daily movement of cell phone users of lower socio-economic status covered greater distances in the Western Area before and after lockdown, thereby showing a greater risk to COVID-19. The research also shows that groups of higher social status decreased mobility significantly after lockdown and did not return to pre-COVID-19 levels, unlike lower-social status groups., Competing Interests: The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article., (© The Author(s) 2023.)

The advent of geolocated information and communication technologies opens the possibility of exploring how people use space in cities, bringing an important new tool for urban scientists and planners, especially for regions where data are scarce or not available. Here we apply a functional network approach to determine land use patterns from mobile phone records. The versatility of the method allows us to run a systematic comparison between Spanish cities of various sizes. The method detects four major land use types that correspond to different temporal patterns. The proportion of these types, their spatial organization and scaling show a strong similarity between all cities that breaks down at a very local scale, where land use mixing is specific to each urban area. Finally, we introduce a model inspired by Schelling's segregation, able to explain and reproduce these results with simple interaction rules between different land uses.

Social network contacts have significant influence on individual travel behavior. However, transport models rarely consider social interaction. One of the reasons is the difficulty to properly model social influence based on the limited data available. Non-conventional, passively collected data sources, such as Twitter, Facebook or mobile phones, provide large amounts of data containing both social interaction and spatiotemporal information. The analysis of such data opens an opportunity to better understand the influence of social networks on travel behavior. The main objective of this paper is to examine the relationship between travel behavior and social networks using mobile phone data. A huge dataset containing billions of registers has been used for this study. The paper analyzes the nature of co-location events and frequent locations shared by social network contacts, aiming not only to provide understanding on why users share certain locations, but also to quantify the degree in which the different types of locations are shared. Locations have been classified as frequent ( home, work and other) and non-frequent. A novel approach to identify co-location events based on the intersection of users' mobility models has been proposed. Results show that other locations different from home and work are frequently associated to social interaction. Additionally, the importance of non-frequent locations in co-location events is shown. Finally, the potential application of the data analysis results to improve activity-based transport models and assess transport policies is discussed. [ABSTRACT FROM AUTHOR]

In this paper we introduce an advanced platform to label mobile event data with significant subscriber locations in real time. The presented platform is divided into two sections – the learning section and the real-time processing section. During the real-time processing step, we enrich live event streams with significant locations calculated in the learning step using stream and call detail record data. We validate our system by comparing a sample of subscribers' calculated locations with actual locations and give state benchmarks for minimum event counts. The validation confirms that the platform works within desired deviation levels from real locations. The accuracy strongly depends on the event count that we can take into account. Finally, we simulate a real-world scenario and measure the real-time labelling performance of our system. The results of this simulation confirm that our event labelling platform performs sufficiently well to process real event streams for millions of subscribers in real time. [ABSTRACT FROM AUTHOR]