Publications
Impact of Later-Stages COVID-19 Response Measures on Spatiotemporal Mobile Service Usage
The COVID-19 pandemic has affected our lives and how we use network infrastructures in an unprecedented way. While early studies have started shedding light on the link between COVID-19 containment measures and mobile network traffic, we presently lack a clear understanding of the implications of the virus outbreak, and of our reaction to it, on the usage of mobile apps. We contribute to closing this gap, by investigating how the spatiotemporal usage of mobile services has evolved through different response measures enacted in France during a continued seven-month period in 2020 and 2021. Our work complements previous studies in several ways: (i) it delves into individual service dynamics, whereas previous studies have not gone beyond broad service categories; (ii) it encompasses different types of containment strategies, allowing to observe their diverse effects on mobile traffic; (iii) it covers both spatial and temporal behaviors, providing a comprehensive view on the phenomenon. These elements of novelty let us lay new insights on how the demands for hundreds of different mobile services are reacting to the new environment set forth by the pandemics.
DeepRay: Deep Learning Meets Ray-Tracing
Efficient and accurate indoor radio propagation modeling tools are essential for the design and operation of wireless communication systems. Lately, several attempts to combine radio propagation solvers with machine learning (ML) have been made. In this paper, motivated by the recent advances in the area of computer vision, we present a new ML propagation model using convolutional encoder-decoders. Specifically, we couple a ray-tracing simulator with either a U-Net or an SDU-Net, showing that the use of atrous convolutions utilized in SDU-Net can significantly enhance the performance of an ML propagation model. The proposed data-driven framework, called DeepRay, can be trained to predict the received signal strength in a given indoor environment. More importantly, once trained over multiple input geometries, DeepRay can be employed to directly predict the signal level for unknown indoor environments. We demonstrate this approach in various indoor environments using long range (LoRa) devices operating at 868 MHz.
EM DeepRay: An Expedient, Generalizable and Realistic Data-Driven Indoor Propagation Model
Efficient and realistic indoor radio propagation modelling tools are inextricably intertwined with the design and operation of next-generation wireless networks. Machine learning (ML)-based radio propagation models can be trained with simulated or real-world data to provide accurate estimates of the wireless channel characteristics in a computationally efficient way. However, most of the existing research on ML-based propagation models focuses on outdoor propagation modelling, while indoor data-driven propagation models remain site-specific with limited scalability. In this paper, we present an efficient and credible ML-based radio propagation modelling framework for indoor environments. Specifically, we demonstrate how a convolutional encoder-decoder can be trained to replicate the results of a ray-tracer, by encoding physics-based information of an indoor environment, such as the permittivity of the walls, and decoding it as the path-loss (PL) heatmap for an environment of interest. Our model is trained over multiple indoor geometries and frequency bands, and it can eventually predict the PL for unknown indoor geometries and frequency bands within a few milliseconds. Additionally, we illustrate how the concept of transfer learning can be leveraged to calibrate our model by adjusting its pre-estimate weights, allowing it to make predictions that are consistent with measurement data.
Pseudo Ray-Tracing: Deep Leaning Assisted Outdoor mm-Wave Path Loss Prediction
In this letter we present our results on how deep learning can be leveraged for outdoor path loss prediction in the 30GHz band. In particular, we exploit deep learning to boost the performance of outdoor path loss prediction in an end-to-end manner. In contrast to existing 3D ray tracing approaches that use geometrical information to model physical radio propagation phenomena, the proposed deep learning-based approach predicts outdoor path loss in the urban 5G scenario directly. To achieve this, a deep learning model is first trained offline using the data generated from simulations utilizing a 3D ray tracing approach. Our simulation results have revealed that the deep learning based approach can deliver outdoor path loss prediction in the 5G scenario with a performance comparable to a state-of-the-art 3D ray tracing simulator. Furthermore, the deep learning-based approach is 30 times faster than the ray tracing approach.
Forecasting Network Traffic: A Survey and Tutorial With Open-Source Comparative Evaluation
This paper presents a review of the literature on network traffic prediction, while also serving as a tutorial to the topic. We examine works based on autoregressive moving average models, like ARMA, ARIMA and SARIMA, as well as works based on Artifical Neural Networks approaches, such as RNN, LSTM, GRU, and CNN. In all cases, we provide a complete and self-contained presentation of the mathematical foundations of each technique, which allows the reader to get a full understanding of the operation of the different proposed methods. Further, we perform numerical experiments based on real data sets, which allows comparing the various approaches directly in terms of fitting quality and computational costs. We make our code publicly available, so that readers can readily access a wide range of forecasting tools, and possibly use them as benchmarks for more advanced solutions.
A Joint Optimization Approach for Power Efficient Heterogeneous OFDMA Radio Access Networks
Heterogeneous networks have emerged as a popular solution for accommodating the growing number of connected devices and increasing traffic demands in cellular networks. While offering broader coverage, higher capacity, and lower latency, the escalating energy consumption poses sustainability challenges. In this paper a novel optimization approach for OFDMA heterogeneous networks is proposed to minimize transmission power while respecting individual users throughput constraints. The problem is formulated as a mixed integer geometric program, and optimizes at once multiple system variables such as user association, working bandwidth, and base stations transmission powers. Crucially, the proposed approach becomes a convex optimization problem when user-base station associations are provided. Evaluations in multiple realistic scenarios from the production mobile network of a major European operator and based on precise channel gains and throughput requirements from measured data validate the effectiveness of the proposed approach. Overall, our original solution paves the road for greener connectivity by reducing the energy footprint of heterogeneous mobile networks, hence fostering more sustainable communication systems.
Characterizing 5G Adoption and its Impact on Network Traffic and Mobile Service Consumption
The roll out of 5G, coupled with the traffic monitoring capabilities of modern industry-grade networks, offers an unprecedented opportunity to closely observe the impact that the introduction of a new major wireless technology has on the end users. In this paper, we seize such a unique chance, and carry out a first-of-its-kind in-depth analysis of 5G adoption along spatial, temporal and service dimensions. Leveraging massive measurement data about application-level demands collected in a nationwide 4G/5G network, we characterize the impact of the new technology on when, where and how mobile subscribers consume 5G traffic both in aggregate and for individual types of services. This lets us unveil the overall incidence of 5G in the total mobile network traffic, its spatial and temporal fluctuations, its effect on the way 5G services are consumed, the way individual services and geographical locations contribute to fluctuations in the 5G demand, as well as surprising connections between socioeconomic status of local populations and the way the 5G technology is presently consumed.
Characterizing and Modeling Session-Level Mobile Traffic Demands from Large-Scale Measurements
We analyze 4G and 5G transport-layer sessions generated by a wide range of mobile services at over 282, 000 base stations (BSs) of an operational mobile network, and carry out a statistical characterization of their demand rates, associated traffic volume and temporal duration. Our study unveils previously unobserved session-level behaviors that are specific to individual mobile applications and persistent across space, time and radio access technology. Based on the gained insights, we model the arrival process of sessions at heterogeneously loaded BSs, the distribution of the session-level load and its relationship with the session duration, using simple yet effective mathematical approaches. Our models are fine-tuned to a variety of services, and complement existing tools that mimic packet-level statistics or aggregated spatiotemporal traffic demands at mobile network BSs. They thus offer an original angle to mobile traffic data generation, and support a more credible performance evaluation of solutions for network planning and management. We assess the utility of the models in practical application use cases, demonstrating how they enable a more trustworthy evaluation of solutions for the orchestration of sliced and virtualized networks
Impact of later-stages COVID-19 response measures on spatiotemporal mobile service usage
The COVID-19 pandemic has affected our lives and how we use network infrastructures in an unprecedented way. While early studies have started shedding light on the link between COVID-19 containment measures and mobile network traffic, we presently lack a clear understanding of the implications of the virus outbreak, and of our reaction to it, on the usage of mobile apps. We contribute to closing this gap, by investigating how the spatiotemporal usage of mobile services has evolved through different response measures enacted in France during a continued seven-month period in 2020 and 2021. Our work complements previous studies in several ways: (i) it delves into individual service dynamics, whereas previous studies have not gone beyond broad service categories; (ii) it encompasses different types of containment strategies, allowing to observe their diverse effects on mobile traffic; (iii) it covers both spatial and temporal behaviors, providing a comprehensive view on the phenomenon. These elements of novelty let us lay new insights on how the demands for hundreds of different mobile services are reacting to the new environment set forth by the pandemics.
Impact of Public Protests on Mobile Networks
We propose an analytical framework based on a simple metric and capable of analyzing mobile network data so as to identify changes in consumption patterns across antennas due to the occurrence of massive public protests. We collect data from an operational network in France and analyze how it was impacted by the 2023 French pension reform strikes. We are able to identify a number of antennas that were clearly affected by the strike, and to follow the corresponding events in the mobile traffic demand as it propagates in space and time along the designated route followed by the marchers. The proposed framework is a stepping stone for more robust classification models on the impacts of massive protests on mobile networks, paving the road to network-based solutions for a pervasive and cost-effective monitoring of such events
Spatial and Temporal Exploratory Factor Analysis of Urban Mobile Data Traffic
Mobile data traffic is characterized by complex spatiotemporal fluctuations that are linked in entangled ways to the mobility and diverse activities of the mobile network subscribers. Unraveling such dynamics and understanding their root causes are challenging tasks that call for dedicated, complex data analysis tools. In this paper, we propose to employ Exploratory Factor Analysis (EFA) as a unified approach to identify both spatial and temporal structures hidden in the mobile data traffic. We provide a brief introduction to the EFA methodology, discuss how it can be tailored to a networking context, and outline its advantages in terms of versatility, unsupervised nature and interpretability of results. Experiments with large-scale measurement data collected in two urban regions demonstrate the effectiveness of the approach, which allows recognizing and explaining a variety of fundamental structures that underpin real-world spatiotemporal traffic dynamics. A thorough discussion of the results provides interesting insights, including that a reasonably small number of latent factors can describe well the majority of temporal and spatial structures observed in mobile traffic demands, providing valuable insights into key spatiotemporal patterns of population and becoming a valuable asset in understanding the attractiveness factors in urban areas.
AI-assisted Indoor Wireless Network Planning with Data-Driven Propagation Models
Propelled by rapid advances in artificial intelligence (AI), the design and operation of 5G and beyond networks are anticipated to be radically different from those of legacy communication systems. Indeed, AI can be exploited to automate and optimize various essential functionalities of the wireless ecosystem, such as resource allocation, channel modeling, or network planning. This article explores how AI-driven propagation models can be leveraged for the automated and expedient deployment of small cells in indoor environments. To this end, we couple a generalizable data-driven propagation model with an AI-based optimizer to determine the optimal network topology with respect to a target key performance indicator. Our approach reduces the computational time of indoor wireless network design by two to three orders of magnitude, thus enabling accurate planning that would be extremely expensive to conduct using conventional indoor propagation tools and yielding significant gains in the resulting indoor planning quality and performance.
Characterizing Mobile Service Demands at Indoor Cellular Networks
Indoor cellular networks (ICNs) are anticipated to become a principal component of 5G and beyond systems. ICNs aim at extending network coverage and enhancing users’ quality of service and experience, consequently producing a substantial volume of traffic in the coming years. Despite the increasing importance that ICNs will have in cellular deployments, there is nowadays little understanding of the type of traffic demands that they serve. Our work contributes to closing that gap, by providing a first characterization of the usage of mobile services across more than 4, 500 cellular antennas deployed at over 1, 000 indoor locations in a whole country. Our analysis reveals that ICNs inherently manifest a limited set of mobile application utilization profiles, which are not present in conventional outdoor macro base stations (BSs). We interpret the indoor traffic profiles via explainable machine learning techniques, and show how they are correlated to the indoor environment. Our findings show how indoor cellular demands are strongly dependent on the nature of the deployment location, which allows anticipating the type of demands that indoor 5G networks will have to serve and paves the way for their efficient planning and dimensioning.
Stochastic Evaluation of Indoor Wireless Network Performance with Data-Driven Propagation Models
Cell densification through the installation of smallcells and femtocells in indoor environments is an emerging solution to enhance the operation of wireless networks. The deployment of new components within the heart of the radio access network calls for expedient tools that assist and ensure their optimal placement within the existing network infrastructure. In this paper, we introduce metrics that can characterize indoor wireless network performance (IWNP) in terms of coverage and capacity, and we evaluate them via physics-based propagation models. In particular, we exploit a deterministic propagation model, i.e., a ray-tracer, as well as a novel machine learning-based propagation model. We demonstrate that data-driven propagation models can be leveraged for the rigorous evaluation of the IWNP metrics, yielding a remarkable computational efficiency compared to the conventional deterministic models. The use of physics-based sitespecific propagation models allows for the particularities of each indoor geometry to be taken into account, and also makes feasible the consideration of uncertainties related to the indoor environment. In this case, the IWNP metrics are expressed as stochastic quantities and a stochastic solution is derived through an efficient polynomial chaos expansion representation, enabling on-the-fly computation of the IWNP metrics statistics
EM DeepRay: An Expedient, Generalizable and Realistic Data-Driven Indoor Propagation Model
Efficient and realistic indoor radio propagation modelling tools are inextricably intertwined with the design and operation of next generation wireless networks. Machine learning (ML)-based radio propagation models can be trained with simulated or real-world data to provide accurate estimates of the wireless channel characteristics in a computationally efficient way. However, most of the existing research on ML-based propagation models focuses on outdoor propagation modelling, while indoor data-driven propagation models remain site-specific with limited scalability. In this paper we present an efficient and credible ML-based radio propagation modelling framework for indoor environments. Specifically, we demonstrate how a convolutional encoder-decoder can be trained to replicate the results of a ray-tracer, by encoding physics-based information of an indoor environment, such as the permittivity of the walls, and decode it as the path-loss (PL) heatmap for an environment of interest. Our model is trained over multiple indoor geometries and frequency bands, and it can eventually predict the PL for unknown indoor geometries and frequency bands within a few milliseconds. Additionally, we illustrate how the concept of transfer learning can be leveraged to calibrate our model by adjusting its pre-estimate weights, allowing it to make predictions that are consistent with measurement data
DeepRay: Deep Learning Meets Ray Tracing
Efficient and accurate indoor radio propagation modeling tools are essential for the design and operation of wireless communication systems. Lately, several attempts to combine radio propagation solvers with machine learning (ML) have been made. In this paper, motivated by the recent advances in the area of computer vision, we present a new ML propagation model using convolutional encoder-decoders. Specifically, we couple a ray-tracing simulator with either a U-Net or an SDU-Net, showing that the use of atrous convolutions utilized in SDU-Net can significantly enhance the performance of an ML propagation model. The proposed data-driven framework, called DeepRay, can be trained to predict the received signal strength in a given indoor environment. More importantly, once trained over multiple input geometries, DeepRay can be employed to directly predict the signal level for unknown indoor environments. We demonstrate this approach in various indoor environments using long range (LoRa) devices operating at 868 MHz.
Deep Learning-based Multivariate Time Series Classification for Indoor Outdoor Detection
Recently, the topic of indoor outdoor detection (IOD) has seen its popularity increase, as IOD models can be leveraged to augment the performance of numerous Internet of Things and other applications. IOD aims at distinguishing in an efficient manner whether a user resides in an indoor or an outdoor environment, by inspecting the cellular phone sensor recordings. Legacy IOD models attempt to determine a user’s environment by comparing the sensor measurements to some threshold values. However, as we also observe in our experiments, such models exhibit limited scalability, and their accuracy can be poor. Machine learning (ML)-based IOD models aim at removing this limitation, by utilizing a large volume of measurements to train ML algorithms to classify a user’s environment. Yet, in most of the existing research, the temporal dimension of the problem is disregarded. In this paper, we propose treating IOD as a multivariate time series classification (TSC) problem, and we explore the performance of various deep learning (DL) models. We demonstrate that a multivariate TSC approach can be used to monitor a user’s environment, and predict changes in its state, with greater accuracy compared to conventional approaches that ignore the feature variation over time. Additionally, we introduce a new DL model for multivariate TSC, exploiting the concept of self-attention and atrous spatial pyramid pooling. The proposed DL multivariate TSC framework exploits only low power consumption sensors to infer a user’s environment, and it outperforms state-of-the-art models, yielding a higher accuracy combined with a smaller computational cost
Henna: hierarchical machine learning inference in programmable switches
The recent proliferation of programmable network equipment has opened up new possibilities for embedding intelligence into the data plane. Deploying models directly in the data plane promises to achieve high throughput and low latency inference capabilities that cannot be attained with traditional closed loops involving controlplane operations. Recent efforts have paved the way for the integration of trained machine learning models in resource-constrained programmable switches, yet current solutions have significant limitations that translate into performance barriers when coping with complex inference tasks. In this paper, we present Henna, a first in-switch implementation of a hierarchical classification system. The concept underpinning our solution is that of splitting a difficult classification task into easier cascaded decisions, which can then be addressed with separated and resource-efficient tree-based classifiers. We propose a design of Henna that aligns with the internal organization of the Protocol Independent Switch Architecture (PISA), and integrates state-of-the-art strategies for mapping decision trees to switch hardware. We then implement Henna into a real testbed with off-the-shelf Intel Tofino programmable switches using the P4 language. Experiments with a complex 21-category classification task based on measurement data demonstrate how Henna improves the F1 score of an advanced single-stage model by 21%, while keeping usage of switch resources at 8% on average
Flowrest: Practical Flow-Level Inference in Programmable Switches with Random Forests
User-plane machine learning facilitates low-latency, high-throughput inference at line rate. Yet, user planes are highly constrained environments, and restrictions are especially marked in programmable switches with limited memory and minimum support for mathematical operations or data types. Thus, current solutions for in-switch inference that are compatible with production-level hardware lack support for complex features or suffer from limited scalability, and hit performance barriers in complex tasks involving large decision spaces. To address this limitation, we present Flowrest, a first complete Random Forest (RF) model implementation that operates at the level of individual flows in commercial switches. Our solution builds on (i) an original framework to embed flow-level machine learning models into programmable switch ASICs, and (ii) novel guidelines for tailoring RF models to operations in programmable switches already at the design stage. We implement Flowrest as an open-source software using the P4 language, and assess its performance in an experimental platform based on Intel Tofino switches. Tests with tasks of unprecedented complexity show how our model can improve accuracy by up to 39% over previous approaches to implement RF models in real-world equipment.
Demonstrating Flow-Level In-Switch Inference
Existing approaches for in-switch inference with Random Forest (RF) models that can run on production-level hardware do not support flow-level features and have limited scalability to the task size. This leads to performance barriers when tackling complex inference problems with sizable decision spaces. Flowrest is a complete RF model framework that fills existing gaps in the existing literature and enables practical flowlevel inference in commercial programmable switches. In this demonstration, we exhibit how Flowrest can classify individual traffic flows at line rate in an experimental platform based on Intel Tofino switches. To this end, we run experiments with real-world measurement data, and show how Flowrest yields improvements in accuracy with respect to solutions that are limited to packet-level inference in programmable hardware.
Showcasing In-Switch Machine Learning Inference
Recent endeavours have enabled the integration of trained machine learning models like Random Forests in resource-constrained programmable switches for line rate inference. In this work, we first show how packet-level information can be used to classify individual packets in production-level hardware with very low latency. We then demonstrate how the newly proposed Flowrest framework improves classification performance relative to the packet-level approach by exploiting flow-level statistics to instead classify traffic flows entirely within the switch without considerably increasing latency. We conduct experiments using measurement data in a real-world testbed with an Intel Tofino switch and shed light on how Flowrest achieves an F1-score of 99% in a service classification use case, outperforming its packet-level counterpart by 8%.
Fast Detection of Cyberattacks on the Metaverse through User-plane Inference
The metaverse is envisioned as a digital world where people can experience an immersive three-dimensional Internet, thanks to the profound integration of different technologies like the Internet of Things (IoT), augmented and virtual reality. From a technical point of view, developing a system of such an unprecedented scale and complexity also opens new challenges in security: a prominent one is the capability to detect and respond to cyberattacks in the shortest time possible, so as not to disrupt the live user experience. In this paper, we discuss how recent advances in user-plane inference can be leveraged to identify malicious traffic generated by IoT devices connected to the metaverse at line rate, ensuring a faster reaction than stateof-the-art approaches where the attack detection is performed in the control plane. We demonstrate the viability of the solution in a programmable network testbed composed of off-the-shelf Intel Tofino switches and with real-world traffic hiding a number of different IoT-based cyberattacks. Our experimental results show that Random Forest models implemented in programmable switches can achieve up to 99% accuracy while using less than 5% of the hardware resources on average in the target case study. Moreover, they quantify the existing trade-off between attack detection precision and user plane resource consumption.
Encrypted Traffic Classification at Line Rate in Programmable Switches with Machine Learning
Encrypted Traffic Classification (ETC) has become an important area of research with Machine Learning (ML) methods being the state-of-the-art. However, most existing solutions either rely on offline ETC based on collected network data or on online ETC with models running in the control plane of Software-Defined Networks (SDN), all of which do not run at line rate and would not meet latency requirements of timesensitive applications in modern networks. This work leverages recent advances in data plane programmability to achieve realtime ETC in programmable switches at line rate, with high throughput and low latency. The proposed solution comprises (i) an ETC-aware Random Forest (RF) modelling process where only features based on packet size and packet arrival times are used, and (ii) an encoding of the trained RF model into production-grade P4-programmable switches. The performance of the proposed in-switch ETC framework is evaluated using 3 encrypted traffic datasets with experiments in a real-world testbed with Intel Tofino switches, in the presence of background traffic at 40 Gbps. Results show how the solution achieves high classification accuracy of up to 95%, with sub-microsecond delay, while consuming on average less than 10% of total available switch hardware resources.
Towards Data-Driven Management of Mobile Networks through User Plane Inference
Growing network complexity has rendered humanin-the-loop network management approaches obsolete. The advent of Software-Defined Networking (SDN) has enabled network automation, with Machine Learning (ML) models running in the control plane. However, such control plane models do not run at line rate and would not satisfy the stringent latency requirements of time-sensitive next-generation applications. In this PhD project, we exploit recent advances in programmable switches and associated languages like P4 to enable data-driven management of networks by running ML models for inference in programmable switches at line rate, with high throughput and low latency. Resulting contributions include solutions for in-switch classification at packet level, flow level, or both, with use cases in network security, service identification, and device fingerprinting in commercial off-the-shelf switches.
Deep Learning-Based Path Loss Prediction For Outdoor Wireless Communication Systems
Deep learning (DL) has been recently leveraged for the inference of characteristics related to wireless communication channels, such as path loss (PL). This paper presents how a deep convolutional encoder-decoder, namely a path loss prediction net (PPNet) based on SegNet, can be trained to transform information related to an outdoor propagation environment into a PL heatmap. This work is a part of the 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing First Pathloss Radio Map Prediction Challenge. The DL model is trained with synthetic data generated with a high-performance ray tracing simulator and it is illustrated that PPNet can indeed learn to predict the PL distribution and that it generalizes well to previously unseen outdoor propagation environments.
Pseudo Ray-Tracing: Deep Leaning Assisted Outdoor mm-Wave Path Loss Prediction
In this letter we present our results on how deep learning can be leveraged for outdoor path loss prediction in the 30GHz band. In particular, we exploit deep learning to boost the performance of outdoor path loss prediction in an end-to-end manner. In contrast to existing 3D ray tracing approaches that use geometrical information to model physical radio propagation phenomena, the proposed deep learning-based approach predicts outdoor path loss in the urban 5G scenario directly. To achieve this, a deep learning model is first trained offline using the data generated from simulations utilizing a 3D ray tracing approach. Our simulation results have revealed that the deep learning based approach can deliver outdoor path loss prediction in the 5G scenario with a performance comparable to a state-of-the-art 3D ray tracing simulator. Furthermore, the deep learning-based approach is 30 times faster than the ray tracing approach.
IRDM: A generative diffusion model for indoor radio map interpolation
This article proposes a novel methodology for interpolating path-loss radio maps, which are vital for comprehending signal distribution and hence planning indoor wireless networks. The approach employs generative diffusion models and proves to be highly effective in generating accurate radio maps with only a small number of measurements. The experimental outcomes demonstrate an average root-mean-square error of 4.23 dB using only 10 percent of the reference points, highlighting the ability of the generative diffusion model to achieve significant interpolation accuracy in radio map generation.