IMDEA Networks Institute Digital Repository
Not a member yet
1915 research outputs found
Sort by
Exploring Online Manifestations of Real-World Inequalities
Get PDFSocioeconomic gaps, particularly income inequality, affect crime and public opinion. Although official data sources can identify these patterns of income-based social disparity, a fundamental question remains: Can similar social inequalities be found using abundant Internet user activity? (i) How does a neighbourhood's income affect crime discussion in the neighbourhood? (ii) Can user-generated data predict a neighbourhood's income? To answer these questions, we collected 2.5 million Nextdoor posts from 67608 US and UK neighbourhoods between November 2020 and September 2021. We use official US and UK data sources for crime and income information.TRUEpu
Online Learning for Industrial IoT: The Online Convex Optimization Perspective
Get PDFIndustrial Internet of things (IIoT), one enabler for Industry 4.0 Smart Factories, is a mission-critical and latency-sensitive application of 5G networks. Due to the stringent latency requirements in IIoT, coordinating the simultaneous transmissions of massive entities and knowing the interference they create to each other is not feasible. Additionally, due to the mobility feature of mobile robots and automated guided vehicles, the experienced channel fading may differ from the estimated one. Therefore, some uncertainties exist in IIoT networks while we decide the communication and control mechanisms. Within the context of IIoT, this paper discusses some resource allocation solutions from the perspective of Online Convex Optimization (OCO). OCO is a computationally lightweight and memory-efficient mathematical tool which tackles the optimization problems, given that the network environment is arbitrary and unknown. We first introduce the key performance indicators in IIoT networks and highlight the uncertain factors, which we may encounter while allocating the communication resources in IIoT. Then we provide an overview of main principles of OCO and present the comparison benchmarks and related metrics for performance evaluation. Moreover, we discuss the kind of resource allocation problems in IIoT that can be tackled by OCO. Finally, we summarize the advantages of applying OCO to IIoT networks.This work was supported by the CHIST-ERA grant CHIST-ERA-18-SDCDN-004 (grant number T11EPA4- 00056) through the General Secretariat for Research and Innovation (GSRI).TRUEpu
Second-level Digital Divide: a Longitudinal Study of Mobile Traffic Consumption Imbalance in France
Get PDFWe study the interaction between the consumption of digital services via mobile devices and urbanization levels, using measurement data collected in an operational network serving the whole territory of France. We unveil that such an interaction follows a power law, or, in other words, there exists an emergent behavior that prompts subscribers living in increasingly extended and populate urban areas to exhibit a surging individual consumption of mobile traffic. The result holds for the global traffic, but is also consistently observed across a range of mobile services, although with varying intensity. An unprecedented longitudinal analysis of the phenomenon unveils how the imbalance in the per-capita mobile data traffic usage across cities of different size has grown steadily and substantially in the 2014–2019 time frame in France. Our study raises questions on the presence of second-level digital divides in developed countries, and paves the road to further investigations.Comunidad de MadridAgence Nationale de la RechercheTRUEpu
Improving Epidemic Risk Maps Using Mobility Information from Mobile Network Data
Get PDFIn this paper we propose a method for using mobile network data to detect potential COVID-19 hospitalizations and derive corresponding epidemic risk maps. We apply our methods to a dataset from more than 2 million cellphones, collected over the months of March and April in 2020 by a British mobile network provider. The method consists of different algorithms, including detection, filtering, validation and fine-tuning. The approach detected over 2,800 potentially hospitalized individuals, yielding a 98.6\% agreement with released public records of patients admitted to NHS hospitals. Analyzing the mobility pattern of these individuals prior to their potential hospitalization, we present a series of risk maps. Compared with census-based maps, our risk maps indicate that the areas of highest risk are not necessarily the most densely populated ones. We also show that the areas of highest risk may change from day to day. Finally, we observe that hospitalized individuals tended to have a higher average mobility than non-hospitalized ones. Overall, we conclude that the rich spatio-temporal information extracted from mobile network data may benefit both the mobile-based technologies and the policies that are being developed against existing and future epidemics.TRUEpu
FL-Torrent: decentralised AI for the masses
Get PDFGoogle has made Federated Learning (FL) readily available to the public with their software and APIs as TensorFlow, but vendor lock-in is still a problem for experimental deployments of Federated Learning. It is therefore necessary to abide by the design of such proprietary APIs, implying full trust in their safety and security against backdoor or label flipping attacks, similarly to what used to occur in proprietary Operating Systems a while ago. More recently, several criticisms about the privacy of novel cookie killer applications of FL such as FL-of-Cohorts (FLoC) has also driven to the discontinuity of this web based system by Google, which just shows how industry solutions are often imperfect naturally. FL-Torrent aims to decentralise and thus democratise this vendor dominated API ecosystem for Federated Learning.FALSEpu
Forecasting for Network Management with Joint Statistical Modelling and Machine Learning
Get PDFForecasting is a task of ever increasing importance for the operation of mobile networks, where it supports anticipatory decisions by network intelligence and enables emerging zero-touch service and network management models. While current trends in forecasting for anticipatory networking lean towards the systematic adoption of models that are purely based on deep learning approaches, we pave the way for a different strategy to the design of predictors for mobile network environments. Specifically, following recent advances in time series prediction, we consider a hybrid approach that blends statistical modelling and machine learning by means of a joint training process of the two methods. By tailoring this mixed forecasting engine to the specific requirements of network traffic demands, we develop a Thresholded Exponential Smoothing and Recurrent Neural Network (TES-RNN) model. We experiment with TESRNN in two practical network management use cases, i.e., (i) anticipatory allocation of network resources, and (ii) mobile traffic anomaly prediction. Results obtained with extensive traffic workloads collected in an operational mobile network show that TES-RNN can yield substantial performance gains over current state-of-the-art predictors in both applications considered.TRUEpu
Augmenting mmWave Localization Accuracy Through Sub-6 GHz on Off-the-Shelf Devices
Get PDFMillimeter-wave (mmWave) technology is an important element to increase the throughput and reduce latency of future wireless networks. At the same time, its high bandwidth and highly directional antennas allow for unprecedented accuracy in wireless sensing and localization applications. In this paper, we thoroughly analyze mmWave localization and find that it is either extremely accurate or has a very high error, since there is significant mmWave coverage via reflections and even through walls. As a consequence, sub-6 GHz technology can not only provide (coarse) localization where mmWave is not available, but is also critical to decide among multiple candidate antennas and APs for accurate mmWave localization.
Based on these insights, we design a high-accuracy joint mmWave and sub-6 GHz location system. We enable CSI-based angle estimation and FTM-based ranging on off-the-shelf mmWave devices to implement our mechanism and carry out an extensive measurement campaign. Our system is the first to achieve 18~cm median location error with off-the-shelf devices under their normal mode of operation. We further release the location system (and in particular the CSI and FTM functionality) as well as the trace data from the measurement campaign to the research community.TRUEpu
MIMORPH: A General-Purpose Experimentation Platform for sub-6 GHz and mmWave Frequency Bands
Get PDFWith the rapid increase in performance and complexity of wireless networks, it has become
challenging to build experimentation platforms that can meet such performance requirements
but at the same time are comparatively easy to use and flexible. Te lack of suitable platforms
inspired us to build MIMORPH, a single experimentation platform that supports massive MIMO
sub-6 GHz systems, ultra-high bandwidth Millimeter Wave (mmWave) MIMO, as well as mixed sub-6 GHz
and mmWave confgurations. It can be operated in a closed-loop manner and is intended for WLAN, 5G-NR
and future 6G research. MIMORPH is built on top of standard components, such as a state-of-the-art RFSoC FPGA system and its implementation is made freely available to the research community.Region of Madrid through TAPIR-CM (S2018/TCS-4496)Spanish Ministry of Science and Innovation (MICIU) grant RTI2018-094313-B-I00 (PinPoint5G+)European Union’s Horizon 2020 research and innovation program under Grant No. 871249 (LOCUS)TRUEpu
Using mobile network data to color epidemic risk maps
Get PDFIn this paper we propose a method for using mobile network data to detect potential COVID-19 hospitalizations and derive corresponding epidemic risk maps. We apply our methods to a dataset from more than 2 million cellphones, collected over the months of March and April in 2020 by a British mobile network provider. The method consists of different algorithms, including detection, filtering, validation and fine-tuning. The approach detected over 2,800 potentially hospitalized individuals, yielding a 98.6\% agreement with released public records of patients admitted to NHS hospitals. Analyzing the mobility pattern of these individuals prior to their potential hospitalization, we present a series of risk maps. Compared with census-based maps, our risk maps indicate that the areas of highest risk are not necessarily the most densely populated ones. We also show that the areas of highest risk may change from day to day. Finally, we observe that hospitalized individuals tended to have a higher average mobility than non-hospitalized ones. Overall, we conclude that the rich spatio-temporal information extracted from mobile network data may benefit both the mobile-based technologies and the policies that are being developed against existing and future epidemics.Regional Government of MadridEuropean Regional Development FundTRUEpu
Infrastructure-less D2D Communications through Opportunistic Networks
Get PDFIn 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 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.Telematics EngineeringUniversidad Carlos III de Madrid, Spai