Federated unlearning (FU) algorithms allow clients in federated settings to exercise their "right to be forgotten" by removing the influence of their data from a collaboratively trained model.Existing FU methods maintain data privacy by performing unlearning locally on the client-side and sending targeted updates to the server without exposing forgotten data; yet they often rely on server-side cooperation, revealing the client's intent and identity without enforcement guarantees -compromising autonomy and unlearning privacy.In this work, we propose EFU (Enforced Federated Unlearning), a cryptographically enforced FU framework that enables clients to initiate unlearning while concealing its occurrence from the server.Specifically, EFU leverages functional encryption to bind encrypted updates to specific aggregation functions, ensuring the server can neither perform unauthorized computations nor detect or skip unlearning requests.To further mask behavioral and parameter shifts in the aggregated model, we incorporate auxiliary unlearning losses based on adversarial examples and parameter importance regularization.Extensive experiments show that EFU achieves nearrandom accuracy on forgotten data while maintaining performance comparable to full retraining across datasets and neural architectures -all while concealing unlearning intent from the server.Furthermore, we demonstrate that EFU is agnostic to the underlying unlearning algorithm, enabling secure, function-hiding, and verifiable unlearning for any client-side FU mechanism that issues targeted updates.

Dynamic task offloading is essential for real-time robotic applications, enabling them to adapt to fluctuating computational demands and maintain efficiency under changing conditions. This paper introduces a dynamic task offloading framework that incorporates monitoring, decision making, offloading triggering, and performance monitoring to optimize resource usage by offloading real-time tasks to edge and cloud servers. A use case from the manufacturing industry demonstrates the framework’s application, enhancing robotic functions like motion planning. WebAssembly enables the execution of the robotics application across diverse environments, improving both portability and computational efficiency. By addressing key challenges, this work sets the stage for future offloading frameworks to meet the evolving needs of robotic systems.

Multi-Access Edge Computing (MEC) is a paradigm that enables Internet-of-Things (IoT) applications and devices to run tasks in different locations, from IoT devices to servers in the Cloud. This way, less capable devices can offload computation loads to more powerful or available servers. However, choosing the optimal location for a particular task can be complex due to the features of each location and restrictions of the task. For this, numerous approaches in the literature adopt a centralized strategy for the computation offloading decision, which introduces a single point of failure and can be a bottleneck for resource-demanding applications. In this work, we propose a decentralized Deep Reinforcement Learning (DRL) agent to solve the choice of computing locations, and its assessment in a real testbed. This testbed is formed of an end-user device running the agent, which connects to a MEC server and a Cloud server through 5G. We compare the algorithm against four alternatives, one based on another DRL approach, and analyze their performance in terms of meeting the computing tasks’ requirements and energy consumption in the User Equipment (UE), where synthetically generated tasks are executed or offloaded. DRL algorithms are shown to provide the best tradeoff between performance and energy consumption in changing conditions.

Device heterogeneity poses major challenges in Federated Learning (FL), where resource-constrained clients slow down synchronous schemes that wait for all updates before aggregation. Asynchronous FL addresses this by incorporating updates as they arrive, substantially improving efficiency. While its efficiency gains are well recognized, its privacy costs remain largely unexplored—particularly for high-end devices that contribute updates more frequently, increasing their cumulative privacy exposure. This paper presents the first comprehensive analysis of the efficiency-fairness-privacy trade-off in synchronous vs. asynchronous FL under realistic device heterogeneity. We empirically compareFedAvgand staleness-aware FedAsync using a physical testbed of five edge devices spanning diverse hardware tiers, integrating Local Differential Privacy (LDP) and the Moments Accountant to quantify per-client privacy loss. Using Speech Emotion Recognition (SER) as a privacy-critical benchmark, we show that FedAsync achieves up to 10× faster convergence but exacerbates fairness and privacy disparities: high-end devices contribute 6–10× more updates and incur up to 5× higher privacy loss, while low-end devices suffer amplified accuracy degradation due to infrequent, stale, and noise-perturbed updates. These findings motivate the need for adaptive FL protocols that jointly optimize aggregation and privacy mechanisms based on client capacity and participation dynamics, moving beyond static, one-size-fits-all solutions.

Efficient placement of applications and data is critical in edge computing environments, where user mobility causes previously optimal deployments to become less effective over time, impacting performance and delay constraints. This paper investigates coordinated migration strategies that jointly consider application and data placement to improve service responsiveness and resource efficiency. We propose two Multi-Criteria Decision-Making (MCDM) policies, one using balanced weighting and another optimized via a genetic algorithm, to dynamically select edge servers based on multiple factors such as network distance and server utilization. Through extensive simulation using EdgeSimPy in a federated learning context, our results show that the optimized MCDM policy significantly improves request completion rates and reduces total data transferred, outperforming baseline strategies under various system conditions. These findings highlight the importance of adaptive placement mechanisms in maximizing Quality of Service (QoS) at the edge.

Using machine learning to detect and identify cyberattacks in Industrial Control Systems (ICS) offers a promising solution for uncovering zero-day attacks that traditional rulebased models cannot detect. However, applying ML-based intrusion detection in ICS environments presents challenges, including limited availability of attack data and difficulty in accurately identifying attack types. This paper addresses these challenges by proposing two key strategies. First, we demonstrate that the predictable traffic patterns of ICS networks enable the use of semi-supervised learning models for attack detection. We validate this approach using a benchmark dataset, showing that semi-supervised models achieve comparable performance to fully supervised models while relying solely on training with normal network data. Second, we propose a sequence-based approach for attack identification, using temporal data to improve the accuracy of identifying specific attack types. Our experiments reveal that incorporating historical network parameters improves the attack identification. Our research underscores the potential of semisupervised learning for effective attack detection and highlights the importance of incorporating network temporal properties to improve attack identification.

Federated Learning (FL) enables model training across decentralized devices by communicating solely local model updates to an aggregation server. Although such limited data sharing makes FL more secure than centralized approached, FL remains vulnerable to inference attacks during model update transmissions. Existing secure aggregation approaches rely on differential privacy or cryptographic schemes like Functional Encryption (FE) to safeguard individual client data. However, such strategies can reduce performance or introduce unacceptable computational and communication overheads on clients running on edge devices with limited resources. In this work, we present EncCluster , a novel method that integrates model compression through weight clustering with recent decentralized FE and privacy-enhancing data encoding using probabilistic filters to deliver strong privacy guarantees in FL without affecting model performance or adding unnecessary burdens to clients. We performed a comprehensive evaluation, spanning various datasets and architectures, to demonstrate EncCluster scalability across encryption levels. Our findings reveal that EncCluster significantly reduces communication costs — below even conventional FedAvg — and accelerates encryption by more than four times over all baselines; at the same time, it maintains high model accuracy and enhanced privacy assurances.

This paper presents a cybersecurity solution designed to fortify Industrial Control Systems (ICS) against cyberattacks. The proposed solution integrates a Network-based Intrusion Detection System (NIDS) with a Decision Support System (DSS), leveraging machine learning to detect anomalies in network data and employing a filtering mechanism to reduce false alarms. The NIDS protects a simulated ICS testbed, detecting anomalies and forwarding them to the DSS for further analysis and selection of mitigation strategies. We outline the system architecture and showcase promising outcomes from a prototype implementation. Our proof of concept evaluation demonstrates high accuracy in detecting attack scenarios. Challenges such as detection delays between attacks and potential mitigations high-light areas for future improvement. This research contributes to bridging the gap between ML-based IDS and security solutions, paving the way for enhanced cybersecurity in ICS environments.

Federated learning (FL) as a novel paradigm in Artificial Intelligence (AI), ensures enhanced privacy by eliminating data centralization and brings learning directly to the edge of the user's device. Nevertheless, new privacy issues have been raised particularly during training and the exchange of parameters between servers and clients. While several privacy-preserving FL solutions have been developed to mitigate potential breaches in FL architectures, their integration poses its own set of challenges. Incorporating these privacy-preserving mechanisms into FL at the edge computing level can increase both communication and computational overheads, which may, in turn, compromise data utility and learning performance metrics. This paper provides a systematic literature review on essential methods and metrics to support the most appropriate trade-offs between FL privacy and other performance-related application requirements such as accuracy, loss, convergence time, utility, communication, and computation overhead. We aim to provide an extensive overview of recent privacy-preserving mechanisms in FL used across various applications, placing a particular focus on quantitative privacy assessment approaches in FL and the necessity of achieving a balance between privacy and the other requirements of real-world FL applications. This review collects, classifies, and discusses relevant papers in a structured manner, emphasizing challenges, open issues, and promising research directions.

Context: Federated Learning is an approach to distributed machine learning that enables collaborative model training on end devices. FL enhances privacy as devices only share local model parameters instead of raw data with a central server. However, the central server or eavesdroppers could extract sensitive information from these shared parameters. This issue is crucial in applications like speech emotion recognition (SER) that deal with personal voice data. To address this, we propose Optimized Paillier Homomorphic Encryption (OPHE) for SER applications in FL. Paillier homomorphic encryption enables computations on ciphertext, preserving privacy but with high computation and communication overhead. The proposed OPHE method can reduce this overhead by combing Paillier homomorphic encryption with pruning. So, we employ OPHE in one of the use cases of a large research project (DAIS) funded by the European Commission using a public SER dataset.

Modern manufacturing systems collect a huge amount of data which gives an opportunity to apply various Machine Learning (ML) techniques.The focus of this paper is on the detection of anomalous behavior in industrial manufacturing systems by considering the temporal nature of the manufacturing process.Long Short-Term Memory (LSTM) networks are applied on a publicly available dataset called Modular Ice-cream factory Dataset on Anomalies in Sensors (MIDAS), which is created using a simulation of a modular manufacturing system for ice cream production.Two different problems are addressed: anomaly detection and anomaly classification.LSTM performance is analysed in terms of accuracy, execution time, and memory consumption and compared with non-time-series ML algorithms including Logistic Regression, Decision Tree, Random Forest, and Multi-Layer Perceptron.The experiments demonstrate the importance of considering the temporal nature of the manufacturing process in detecting anomalous behavior and the superiority in accuracy of LSTM over non-time-series ML algorithms.Additionally, runtime adaptation of the predictions produced by LSTM is proposed to enhance its applicability in a real system.

Context: Speech Emotion Recognition (SER) is a valuable technology that identifies human emotions from spoken language, enabling the development of context-aware and personalized intelligent systems.To protect user privacy, Federated Learning (FL) has been introduced, enabling local training of models on user devices.However, FL raises concerns about the potential exposure of sensitive information from local model parameters, which is especially critical in applications like SER that involve personal voice data.Local Differential Privacy (LDP) has prevented privacy leaks in image and video data.However, it encounters notable accuracy degradation when applied to speech data, especially in the presence of high noise levels.In this paper, we propose an approach called LDP-FL with CSS, which combines LDP with a novel client selection strategy (CSS).By leveraging CSS, we aim to improve the representatives of updates and mitigate the adverse effects of noise on SER accuracy while ensuring client privacy through LDP.Furthermore, we conducted model inversion attacks to evaluate the robustness of LDP-FL in preserving privacy.These attacks involved an adversary attempting to reconstruct individuals' voice samples using the output labels provided by the SER model.The evaluation results reveal that LDP-FL with CSS achieved an accuracy of 65-70%, which is 4% lower than the initial SER model accuracy.Furthermore, LDP-FL demonstrated exceptional resilience against model inversion attacks, outperforming the non-LDP method by a factor of 10.Overall, our analysis emphasizes the importance of achieving a balance between privacy and accuracy in accordance with the requirements of the SER application.

Heterogeneous networks (hetnets) is an interconnection of distinctive networking paradigms to enable wider reachability and greater collaborations. In large Internet-of-Things (IoT) applications, many wireless networks are spatially co-located and intertwined forming hetnets; for instance, health monitoring devices utilising ZigBee or IEEE 802.15.4 co-exist in 2.4 GHz spectrum alongside Wi-Fi devices. Interoperability or non-obtrusive operations are required among the disjoint domains to achieve operational efficiency in overall IoT ecosystem. Specifically, network interoperability in hetnets assure desired reachability, resource orchestration and network quality. In this work, we have modelled and implemented a simulation environment for hetnets to support different schemes of network interoperability under distributed and centralised management of network. The implementation has been evaluated for network scalability and reliability to replicate large IoT hetnets. By evaluating against increasing number of nodes in the hetnet, the mean latency under distributed management is improved by 100-fold with the centralised management. Similar observations could also be made for throughput and packet loss rate.

Many Intelligent Transportation System (ITS) applications rely on communication between fixed ITS stations (roadside installations) and mobile ITS stations (vehicles) to provide traffic safety. In VANETs, the Control Channel (CCH) and Service Channels (SCHs) are applied to transmit the safetyrelated data. The CCH used in IEEE 802. 11p standard for exchanging high-priority safety messages and control information and it can be easily congested by high-frequency periodic beacons under high density scenarios. Also, employing the DedicatedShort Range Communication (DSRC) band of IEEE 802.11p standard can hardly satisfy the requirements of high-critical safety applications. Also, transmitting safety beacons at a constant rate regardless of considering the condition of the links leads to the lack of flexibility and medium resources to support meeting the reliability requirements of these applications. In this paper, we evaluate a beacon rate control method, which assigns a higher beacon rate to nodes based on link conditions, i.e. with more surrounding nodes and better conditions to disseminate beacons. On the other hand, as IEEE 802.11p Medium Access Control (MAC) layer does not perform well under high channel load, so in this paper, we use Self-organizing Time Division Multiple Access (STDMA) in our simulations as MAC layer protocol. The results of the simulations demonstrate the rate of beacon transmission/reception effectively improves, results in better resource utilization. Also, Packet Error Rate (PER) and Packet Inter-Reception time (PIR) decrease significantly which is crucial for safety applications.

Nowadays, road traffic management is becoming a major challenge for realistic society. As a result, reliable communication between vehicles is the key point to this challenge. Currently, IEEE802.11p which is considered as de facto standard for road communication is designed to solve this challenge. However, the communication channel medium is still expected to get congested when a large number of vehicles exist. Target to solve this, European Telecommunication Standards Institute (ETSI) has standardized a set of Decentralized Congestion Control (DCC) mechanisms to control channel load. One of the main topics is achieving channel load control to guarantee reliable communication for platooning systems. In this paper, we focus on investigating on DCC reactive control approaches, aiming to provide comprehensive insights of how DCC framework transmission parameters, i.e. message generation rate, transmission power and data rate, will impact the stability of platooning systems. Besides, for each instance of the transmission parameter, we target to optimize the parameter and propose more stable control algorithms by running repetitive simulations.

Platooning is anticipated to facilitate automated driving even with semi-automated vehicles, by forming road trains using breadcrumb tracing and Cooperative Adaptive Cruise Control (CACC). With CACC, the vehicles coordinate and adapt their speed based on wireless communications. To keep the platoon fuel-efficient, the inter-vehicle distances need to be quite short, which requires automated emergency braking capabilities. In this paper, we propose synchronized braking, which can be used together with existing CACC controllers. In synchronized braking, the leading vehicle in the platoon does not brake immediately, but instead communicates its intentions and then, slightly later, the whole platoon brakes simultaneously. We show that synchronized braking can avoid rear-end collisions even at a very high deceleration rate and with short inter- vehicle distances. Also, the extra distance travelled during the delay before braking can be compensated by enabling a higher deceleration, through coordinated synchronized braking.

In this paper, we study the problem of how to provide reliable communications for energy harvesting (EH) wireless sensor network (WSN). Using the example of an autonomous quarry, where self-driving trucks autonomously collect and transport goods, there is a need for multiple wireless sensors collecting data about where and when goods can be collected, while guaranteeing reliable operation of the quarry. The vehicles transfer energy to the wireless sensors within range, forming a cluster. The sensors use this energy to transmit data to the vehicles. Finally, the vehicles relay information to an access point (AP). The AP processes the collected information and synchronize the operation of all vehicles. We propose an interference channel selection policy for the sensors-to-vehicles links and vehicles-to-AP links to improve the reliability of the communications, while enhancing the energy utilization. Accordingly, closed-form expression on how to achieve reliable communication within the considered system is derived and numerical results show that the proposed channel selection strategy not only improves the probability of achieving sufficiently reliable communication but also enhances the energy utilization.

Industry 4.0 is the German vision for the future of manufacturing, where smart factories use information and communication technologies to digitise their processes to achieve improved quality, lower costs, and increased efficiency. It is likely to bring a massive change to the way control systems function today. Future distributed control systems are expected to have an increased connectivity to the Internet, in order to capitalize on new offers and research findings related to digitalization, such as cloud, big data, and machine learning. A key technology in the realization of distributed control systems is middleware, which is usually described as a reusable software layer between operating system and distributed applications. Various middleware technologies have been proposed to facilitate communication in industrial control systems and hide the heterogeneity amongst the subsystems, such as OPC UA, DDS, and RT-CORBA. These technologies can significantly simplify the system design and integration of devices despite their heterogeneity. However, each of these technologies has its own characteristics that may work better for particular applications. Selection of the best middleware for a specific application is a critical issue for system designers. In this paper, we conduct a survey on available standard middleware technologies, including OPC UA, DDS, and RT-CORBA, and show new trends for different industrial domains.

Safety applications developed for vehicular environments require every vehicle to periodically broadcast its status information (beacon) to all other vehicles, thereby avoiding the risk of car accidents in the road. Due to the high requirements on timing and reliability posed by traffic safety applications, the current IEEE 802.11p standard, which uses a random access Medium Access Control (MAC) protocol, faces difficulties to support timely and reliable data dissemination in vehicular environments where no acknowledgement or RTS/CTS (Request-to-Send/Clear-to-Send) mechanisms are adopted. In this paper, we propose the Dynamic Token-Based MAC (DTB-MAC) protocol. It implements a token passing approach on top of a random access MAC protocol to prevent channel contention as much as possible, thereby improving the reliability of safety message transmissions. Our proposed protocol selects one of the neighbouring nodes as the next transmitter; this selection accounts for the need to avoid beacon lifetime expiration. Therefore, it automatically offers retransmission opportunities to allow vehicles to successfully transmit their beacons before the next beacon is generated whenever time and bandwidth are available. Based on simulation experiments, we show that the DTB-MAC protocol can achieve better performance than IEEE 802.11p in terms of channel utilization and beacon delivery ratio for urban scenarios.

Most applications developed for vehicular environments rely on broadcasting as the main mechanism to disseminate their messages. However, in IEEE 802.11p, which is the most widely accepted Medium Access Control (MAC) protocol for vehicular communications, all transmissions remain unacknowledged if broadcasting is used. Furthermore, safety message transmission requires a strict delay limit and a high reliability, which is an issue for random access MAC protocols like IEEE 802.11p. Therefore, transmission reliability becomes the most important issue for broadcast-based services in vehicular environments. In this paper, we propose a hybrid MAC protocol, referred as Dynamic Token-Based MAC Protocol (DTB-MAC). DTB-MAC uses both a token passing mechanism and a random access MAC protocol to prevent channel contention as much as possible, and to improve the reliability of safety message transmissions. Our proposed protocol tries to select the best neighbouring node as the next transmitter, and when it is not possible, or when it causes a high overhead, the random access MAC protocol is used instead. Based on simulation experiments, we show that the DTB-MAC protocol can achieve better performance compared with IEEE 802.11p in terms of channel utilization and beacon delivery ratio.

Platooning is both a challenging and rewarding application. Challenging since strict timing and reliability requirements are imposed by the distributed control system required to operate the platoon. Rewarding since considerable fuel reductions are possible. As platooning takes place in a vehicular ad hoc network, the use of IEEE 802.11p is close to mandatory. However, the 802.11p medium access method suffers from packet collisions and random delays. Most ongoing research suggests using TDMA on top of 802.11p as a potential remedy. However, TDMA requires synchronization and is not very flexible if the beacon frequency needs to be updated, the number of platoon members changes, or if retransmissions for increased reliability are required. We therefore suggest a token-passing medium access method where the next token holder is selected based on beacon data age. This has the advantage of allowing beacons to be re-broadcasted in each beacon interval whenever time and bandwidth are available. We show that our token-based method is able to reduce the data age and considerably increase reliability compared to pure 802.11p.