Federated learning (FL) enables vehicles to train machine learning (ML) models in a distributed manner for intelligent transportation system (ITS) services without sharing raw data, thereby preserving privacy and reducing communication overhead on roadside units (RSUs). However, due to their mobility and limited communication resources, vehicles participating in FL may be unable to transmit their updated models after local training if they move out of the RSU’s communication range. This limitation not only increases the convergence time of the global model but also wastes the computing resources of the participating vehicles used for training, as their trained models cannot be used to update the global model. To address these issues, this paper proposes a vehicle-to-vehicle communication-assisted FL (V2V-FL) mechanism for vehicular networks, in which vehicles moving out of the RSU’s coverage transmit their updated models to the RSU via a V2V link with limited communication resources, ensuring that more participating vehicles contribute their updated models to FL. An optimization problem for updated model forwarding with resource allocation is formulated to maximize the completion probability of participating vehicles while minimizing the total delay incurred during the training process. Since the formulated problem is a mixed-integer non-convex problem, it is decomposed into two subproblems, which are solved using the Karush-Kuhn-Tucker (KKT) conditions and a heuristic algorithm. Simulation results show that the proposed scheme outperforms other benchmark methods in terms of the total delay and convergence performance.

Vertical Federated Learning (VFL) presents a promising approach for training machine learning models across wearable Internet of Things (IoT) devices while keeping the data local. However, non-Independent and Identically Distributed (IID) data from heterogeneous IoT devices lead to degraded performance of Wearable Human Activity Recognition (WHAR) models due to overfitting. To address non-IID data, this paper introduces the Federated-Multimodal GAN (FM-GAN) approach, leveraging Generative Adversarial Networks (GANs) models to generate an IID distribution of data for balancing multi-modal data.

Unmanned Aerial Vehicles (UAVs) provide an effective solution for supporting Federated Learning (FL) in remote areas lacking infrastructure, but the non-IID nature of data affects FL efficiency. We propose a hierarchical clustering method with pipelined execution for UAV-assisted FL, clustering IoT devices based on computational capacity and data similarity to ensure balanced participation and faster convergence. Simulation results demonstrate improved learning efficiency and stability in non-IID environments, validating the method's effectiveness.

This paper proposes a strategy for offloading unmanned aerial vehicles (UAVs) tasks to electric vehicles and edge servers using a Stackelberg game model. We introduce a suitability score that considers distance, resource availability, and battery level to select the most suitable nodes for offloading UAV tasks. Simulation results demonstrate that the proposed scheme outperforms existing methods in terms of energy consumption and utility for UAVs.

Researchers have studied Internet censorship for nearly as long as attempts to censor contents have taken place. Most studies have however been limited to a short period of time and / or a few countries; the few exceptions have traded off detail for breadth of coverage. Collecting enough data for a comprehensive, global, longitudinal perspective remains challenging.In this work, we present ICLab, an Internet measurement platform specialized for censorship research. It achieves a new balance between breadth of coverage and detail of measurements, by using commercial VPNs as vantage points distributed around the world. ICLab has been operated continuously since late 2016. It can currently detect DNS manipulation and TCP packet injection, and overt "block pages" however they are delivered. ICLab records and archives raw observations in detail, making retrospective analysis with new techniques possible. At every stage of processing, ICLab seeks to minimize false positives and manual validation.Within 53,906,532 measurements of individual web pages, collected by ICLab in 2017 and 2018, we observe blocking of 3,602 unique URLs in 60 countries. Using this data, we compare how different blocking techniques are deployed in different regions and/or against different types of content. Our longitudinal monitoring pinpoints changes in censorship in India and Turkey concurrent with political shifts, and our clustering techniques discover 48 previously unknown block pages. ICLab's broad and detailed measurements also expose other forms of network interference, such as surveillance and malware injection.

Recent years have seen the Internet become a key vehicle for citizens around the globe to express political opinions and organize protests. This fact has not gone unnoticed, with countries around the world repurposing network management tools (e.g., URL filtering products) and protocols (e.g., BGP, DNS) for censorship. Previous work has focused on identifying how censorship is performed. However, there is no major studies to identify, at a global scale, the networks responsible for performing censorship. Also, repurposing network products for censorship can have unintended international impact, which we refer to as "censorship leakage". While there have been anecdotal reports of censorship leakage, there has yet to be a systematic study of censorship leakage at a global scale.