The rapid evolution of beyond fifth-generation (B5G) and sixth-generation (6G) networks is reshaping mobile edge computing (MEC) to support large-scale, heterogeneous Internet of Things (IoT) deployments and complex cyber-physical systems (CPS). Conventional data-driven intelligence in MEC traditionally relies on centralized learning paradigms that often fail to meet the privacy, latency, scalability, and adaptability requirements in distributed and resource-constrained environments. To address these shortcomings, the objective of our work in this paper is to investigate the paradigm shift toward distributed learning and demonstrate how its co-design with emerging communication and system-level technologies can enable scalable and trustworthy intelligence for next-generation IoT and CPS. Building on this objective, we conduct a systematic survey of recent studies and analyze twelve key enabling technologies, including concept drift adaptation, transformers, TinyML, blockchain, integrated sensing and communication (ISAC), digital twins, explainable AI, federated learning and unlearning, adversarial ML, meta-learning, and multi-armed bandits. The surveyed literature is organized using a unified taxonomy and an integrated conceptual pipeline, which clarifies how these enablers interact across sensing, communication, computation, trust, and adaptation layers of IoT and CPSs. The main outcomes of this study include: (i) a comprehensive taxonomy characterizing enabling technologies for distributed edge intelligence, (ii) a comparative synthesis of representative works highlighting common architectural patterns and evaluation practices, and (iii) the identification of research gaps, critical trade-offs, and open challenges, particularly related to model robustness, energy efficiency, data heterogeneity, and secure real-time inference. Overall, this survey establishes a structured foundation and forward-looking roadmap for designing scalable, privacy-preserving, and intelligent distributed learning systems in future B5G- and 6G-enabled IoT and CPS environments.

Most existing image-based parking space detection and classification methods assume that all training data reside in a single, centralized location—an unrealistic scenario that yields models unable to generalize to new parking lots. Furthermore, privacy concerns prevent lot owners from sharing raw images, limiting collaboration. To overcome these challenges, we present ParkFL, the first federated learning framework for parking space detection and classification that trains models across distributed client sites without exchanging raw image data. Built on NVFlare, ParkFL demonstrates model-agnosticism through evaluation on two deep-learning architectures. We benchmark four federated algorithms—FedAvg, FedProx, FedOpt, and SCAFFOLD—using real-world datasets. Despite training on non-centralized, heterogeneous data, ParkFL achieves 99.5% mAP, matching the accuracy of state-of-the-art centralized models on the same parking lots. When evaluated on images from different parking lots, ParkFL significantly outperforms models trained solely on individual-site data, which achieve 25.6% mAP, even though ParkFL never accesses raw images from other sites. Communication overhead remains below 3% of total training time, demonstrating a scalable, privacy-preserving solution with nearly state-of-the-art performance. We release ParkFL code at https://github.com/ahmedmbakr/ParkFL.

While millimeter-wave (mmWave) communication in emerging Sixth Generation (6G) networks offers high bandwidth for the Internet of Things (IoT), it is highly susceptible to blockages, necessitating intelligent relay positioning. Current static relay selection methods are typically unable to adapt to dynamic blockage conditions in IoT deployments, leading to frequent connectivity outages. In this paper, we address this by introducing Hierarchical Mobile Adaptive Relay Control (H-MARC), a reinforcement-learning framework for intelligent mobile relay positioning in IoT networks. H-MARC decomposes relay positioning into strategic long-term planning and tactical real-time control using Twin Delayed Deep Deterministic (TD3) Policy Gradient algorithms. We further present a context-aware extension of H-MARC, referred to as H-MARC-C, by exploiting WiFi context information for predictive blockage detection through cross-band correlation analysis. Computer-based simulations demonstrate that H-MARC achieves 4.2 bits/s/Hz spectral efficiency with 87% connection reliability, while H-MARC-C attains 4.8 bits/s/Hz with 93% reliability representing 35% and 55% improvements over static methods. The framework reduces blockage adaptation time from 2.3s to 0.3s and achieves 80% higher energy efficiency (4.2×106bits/J) compared to reactive approaches, with 24% faster convergence than flat RL (reinforcement learning) baselines, enabling ultra-reliable communications for demanding IoT applications including industrial automation and smart cities.

The evolution toward 6G wireless networks demands ultra-dense cell-free massive MIMO (Multiple-Input Multiple-Output) systems that can deliver unprecedented connectivity while preserving data privacy and enabling rapid adaptation to dynamic conditions. Current resource allocation approaches rely on centralized deep reinforcement learning frameworks that create scalability bottlenecks, require extensive training periods, and violate emerging privacy regulations through global data aggregation. This paper introduces a Privacy-Preserving Federated Meta-Learning (PP-FML) framework that addresses these fundamental limitations through distributed intelligence and instant adaptation mechanisms. The proposed approach enables each access point to learn optimal resource allocation policies locally while collaboratively improving system-wide performance through cryptographically secure gradient sharing. The meta-learning component provides few-shot adaptation capabilities, allowing networks to respond to new conditions within minutes rather than hours. Comprehensive performance evaluation demonstrates that PP-FML achieves 42.3% sum rate improvement, 28.8% better energy efficiency (15.2 bits/Hz/J), sub-minute adaptation latency (51 seconds), and strong privacy guarantees epsilon 1.0 differential privacy compared to centralized approaches while maintaining complete data privacy and enabling rapid adaptation to changing network conditions. The framework scales linearly to ultra-dense deployments exceeding 300 access points per square kilometer with constant per-node computational complexity, making it suitable for practical 6G network deployment with heterogeneous device populations including mobile users and diverse applications.

The emergence of 6G-enabled Vehicle-to-Everything (V2X) networks has created unprecedented demand for ultra-reliable, low-latency spectrum allocation across heterogeneous entities including vehicles, IoT devices, and industrial systems. Current spectrum allocation methods suffer from exponential computational complexity, extensive information sharing requirements, and poor scalability in dense networks. This paper proposes AMADRL (Attention-based Multi-Agent Deep Reinforcement Learning), a novel framework employing dual critic networks with multi-head self-attention mechanisms for intelligent spectrum allocation. The dual critic architecture resolves individual-collective optimization conflicts through local critics for independent entity optimization and a global critic with attention-based coordination. Our approach significantly reduces information sharing requirements while handling heterogeneous QoS demands across diverse entity types. Comprehensive experimental evaluation comparing AMADRL against state-of-the-art baselines including MADDPG, MAAC, QMIX, attention-based methods (A-DDPG, MHA-DQN), and game-theoretic approaches reveals that AMADRL achieves superior performance across multiple metrics including spectrum utilization efficiency, interference mitigation, and network scalability, while preserving user privacy and satisfying strict latency constraints required by safety-critical and industrial use cases.

This paper proposes a centralized digital twin (DT)–driven framework for end-to-end route optimization in flying ad hoc networks (FANETs) that explicitly accounts for channel dynamics in delay-Doppler (DD) domain and residual unmanned aerial vehicles (UAVs)’ energy states. A DT server maintains a high-fidelity virtual representation of the physical FANET using compact DD channel descriptors and UAVs’ energy information, thereby reducing communication overhead while enabling proactive best route selection. In this context, the UAV routing problem is formulated as a centralized context-aware combinatorial bandit. The DT server will utilize combinatorial neural Thompson sampling (CN-TS) or combinatorial neural upper confidence bound (CN-UCB) algorithms to learn the non-linear relationship between UAVs’ dynamics and the achievable end-to-end transmitted data for selecting the optimal multi-hop route sequentially. Numerical results demonstrate that the proposed DT framework significantly outperforms conventional approaches in highly dynamic FANET scenarios.

Machine learning (ML) anomaly detectors are commonly used to identify cyber-attacks on smart power grids because they can detect new (i.e., zero-day) attacks by classifying deviations from normal patterns as anomalies. Deeplearning-based anomaly detectors offer superior performance but are highly sensitive to the selection of threshold values for defining anomalies. Conversely, traditional (or shallow-based) detectors avoid this threshold sensitivity but often underperform, particularly when dealing with complex interdependent data. Moreover, like all ML models, these detectors are vulnerable to adversarial evasion attacks, where adversaries make small and subtle manipulations to false data to evade detection. To address these issues, we propose a robust hybrid-based anomaly detector that combines the strengths of both deep and shallow-based and is trained using explanations derived from power consumption readings rather than the raw readings themselves. This hybrid approach not only mitigates threshold sensitivity and improves performance but also enhances robustness against white-box evasion attacks. Additionally, we introduce an interpretability method using occlusion sensitivity, which helps explain how a classification decision is made for an input power consumption sample, thereby increasing trust, reliability, and understanding of various attack patterns.

Healthcare is revolutionized by the integration of the Internet of Medical Things (IoMT) and artificial intelligence (AI), enabling real-time patient monitoring, advanced predictive analytics, and personalized treatment plans. However, the existing AI healthcare models are typically trained offline on static datasets, limiting their adaptability to the dynamic nature of health data. This may result in compromising models' accuracy and healthcare decision-making, rendering them obsolete. Moreover, attackers may exploit concept drift by injecting frequent data shifts, which can exhaust healthcare institutions' resources. To address this research gap, we propose a novel adaptive, secure, and efficient concept drift detection framework for healthcare. First, a robust deep learning (DL) model is devised to leverage its high-confidence probability to detect data drift efficiently without relying on labeled data. Then, we propose a customized consortium blockchain network that leverages group signatures to ensure anonymity and unlinkability of patients' health data. It also utilizes a dualledger structure, facilitating a unified drift detection model and enabling authenticated, drift-specific data sharing among medical centers. This design protects against data tampering and falsely claiming drift incidents. Our experiments, conducted on a real health monitoring dataset, show that our concept drift detection approach achieves comparable drift detection performance to the existing methods while reducing the computational time by 52.35%, and achieving an accuracy of 98.43 with our offline model and a 95% accuracy with the online adaptive model.

With the rise of Software-Defined Networking (SDN) for managing traffic and ensuring seamless operations across interconnected devices, challenges arise when SDN controllers share infrastructure with deep learning (DL) workloads. Resource contention between DL training and SDN operations, especially in latency-sensitive IoT environments, can degrade SDN's responsiveness and compromise network performance. Federated Learning (FL) helps address some of these concerns by decentralizing DL training to edge devices, thus reducing data transmission costs and enhancing privacy. Yet, the computational demands of DL training can still interfere with SDN's performance, especially under the continuous data streams characteristic of IoT systems. To mitigate this issue, we propose REDUS (Resampling for Efficient Data Utilization in Smart-Networks), a resampling technique that optimizes DL training by prioritizing misclassified samples and excluding redundant data, inspired by AdaBoost. REDUS reduces the number of training samples per epoch, thereby conserving computational resources, reducing energy consumption, and accelerating convergence without significantly impacting accuracy. Applied within an FL setup, REDUS enhances the efficiency of model training on resource-limited edge devices while maintaining network performance. In this paper, REDUS is evaluated on the CICIoT2023 dataset for IoT attack detection, showing a training time reduction of up to 72.6% with a minimal accuracy loss of only 1.62%, offering a scalable and practical solution for intelligent networks.

Network slicing enables multiple virtual networks on shared 6G infrastructure, but dynamic resource allocation across Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low Latency Communications (URLLC), and Massive Machine Type Communications (mMTC) services remains challenging. We present a hybrid Convolutional Neural Network-Long Short-Term Memory Architecture (CNN-LSTM) framework with service-specific utility functions that optimize resources while ensuring Quality of Service (QoS) guarantees under dynamic conditions. Our approach integrates spatial pattern recognition with temporal prediction, incorporating constraint measurement and lightweight optimization. Experimental results on a testbed with 100 base stations and 10,000 users demonstrate superior performance over state-of-the-art methods. The framework achieves significant improvements in resource utilization, QoS satisfaction, and energy efficiency with real-time inference capability. Convergence analysis validates system stability, confirming practical deployment feasibility for latency-critical 6G applications.

The evolution of 6G wireless networks requires robust and adaptive communication systems that can handle dynamic indoor environments. Accurate prediction of the terahertz (THz) channel is a key enabler of this adaptability, enabling proactive decisions such as beamforming and handover. Because traffic levels (i.e., user densities) fluctuate, the underlying channel statistics change over time, resulting in concept drift that can deteriorate the performance and generalization ability of deep learning (DL) models if they are tested against mismatched conditions (i.e., on a traffic level other than the one used for training). This paper introduces a novel framework for generalizable THz channel prediction, enabled by fusing environmental sensing with AI-assisted wireless communication to mitigate concept drift and maintain generalization. First, we investigate the use of DL-based channel prediction models tailored to specific traffic levels—light, moderate, and dense—and evaluate their performance under mismatched conditions. Results show up to 51% performance deterioration when models are exposed to mismatched traffic scenarios. To mitigate this issue, a traffic-aware channel prediction framework is proposed, comprising three stages: people counting using sensing technologies; quantization of the user count into traffic levels; and dynamic selection of the corresponding DL model. Simulation results demonstrate that integrating accurate sensing technologies, particularly vision-based systems, significantly reduces prediction deterioration to as low as 4%. The proposed framework’s adaptability ensures reliable channel prediction by aligning model selection with real-time traffic conditions, which highlights the potential of fusing environmental sensing with AI-assisted wireless communication to enhance the robustness of future 6G networks.

Electromagnetic interference (EMI) still poses a serious threat to reliable wireless communication, particularly in crowded and hostile electromagnetic environments. While space-time block coding (STBC) and other conventional diversity techniques have been extensively employed to mitigate multipath fading, their ability to mitigate EMI is inherently limited, especially when interference uniformly affects every component of the antenna. The spectral efficiency of Fluid Antenna (FA)-enabled Multiple-Input Multiple-Output (MIMO) systems can be significantly enhanced by employing Index Modulation (IM). However, current FA-enabled IM (FAIM)-aided MIMO systems suffer from considerable performance degradation due to strong spatial correlation in the wireless channel, which is caused by the dense port distribution of the FA. In this paper, we propose an effective approach that integrates a Fluid Antenna System (FAS) with STBC to enhance system robustness in the presence of EMI. Simulation results show that integrating an FAS into a dual-antenna receiver employing STBC significantly improves spectral efficiency and reliability under EMI. At low signal-to-noise ratios (SNRs), the presence of EMI reduces Bit Error Rate (BER) performance by more than 12% when compared to the traditional arrangement without EMI. In contrast to the system affected by EMI without FAS, the BER curve closely resembles the scenario without EMI when FAS selection is used, resulting in a ~10% BER reduction at Eb/N0= 6 dB. Similarly, the achievable rate bridges the performance gap caused by EMI by improving by more than 1.5 bps/Hz over the SNR range. These findings confirm that FAS is effective in reducing EMI and improving communication in unfriendly settings.

6G networks are expected to significantly benefit from advanced wireless local area technologies such as Wireless Gigabit (WiGig), which operates in the 60 GHz frequency band. This band supports extremely high data rates and low latency, making it ideal for next-generation wireless applications such as the metaverse and holograms. However, WiGig signals are highly susceptible to attenuation from physical obstructions, resulting in frequent handovers and connectivity disruptions. Traditional reactive handover mechanisms are often slow due to latency in decision-making and processing overhead. However, proactive handover strategies that leverage channel prediction can enhance network reliability and improve the quality of service. This paper investigates the feasibility of using statistical methods, specifically the auto-regressive integrated moving average (ARIMA) model, to predict the received signal strength indicator (RSSI) in real-world indoor WiGig environments. Our results indicate that ARIMA exhibits poor predictive accuracy, with a root mean square error (RMSE) of 15 dBm, which may trigger inaccurate handover decisions by initiating handovers under strong signal conditions or failing to respond under weak ones. To overcome this shortcoming, we propose an ensemble learning-based channel prediction approach utilizing the random forest (RF) algorithm. Our results show that the RF model significantly outperforms ARIMA by effectively capturing the nonlinear dynamics of real-world indoor WiGig channels. Specifically, the RF model achieves a 90% reduction in both mean absolute error and RMSE, and a 99% reduction in mean squared error, offering a promising solution for robust proactive handover management in 6G networks.

The dynamic nature of user mobility and density in indoor WiGig networks poses a significant challenge to seamless handover, particularly in the 60 GHz band, where small and closely clustered channel gain values hinder effective decision-making. To address this, we propose a generalizable deep reinforcement learning (DRL)-based handover that integrates a novel reward function designed to amplify channel gain differentials, thereby improving the convergence speed and decision accuracy of learning agents. We investigate the performance of state-of-the-art DRL algorithms—Deep Q-Network (DQN), Proximal Policy Optimization (PPO), and Advantage Actor-Critic (A2C)—enhanced through advanced hyperparameter-tuning techniques, including grid search, random search, optuna, and hyperopt. Among these, DQN combined with grid search yields the best overall performance, surpassing A2C by 48% in average rolling reward and achieving 32% faster convergence than PPO.To assess generalization, we evaluate the merged DQN agent trained across varying user density scenarios (1–8 users) against expert agents specialized for individual densities and a high-density-trained agent tested across all scenarios. Our results reveal that the merged agent exhibits robust and consistent performance across all densities, indicating strong generalization capability. In contrast, the high-density agent suffers performance degradation of up to 13% when exposed to unseen scenarios, underscoring its limited adaptability. While expert agents perform optimally within their specific environments, their deployment complexity renders them impractical for real-time systems. These findings highlight the importance of training DRL agents across diverse scenarios to achieve scalable and generalizable handover solutions in dense and dynamic WiGig networks.

Emerging wireless local area networks, such as WiGig that operate in the extremely high-frequency band (60 GHz) hold significant potential for the development of next-generation 6G networks by offering high throughput and low latency. However, the 60 GHz band is prone to severe signal degradation due to channel blockages, leading to frequent handovers and challenges in maintaining seamless connectivity. Reactive handover strategies can result in service delays due to overhead and decision-making latency. To tackle these issues, proactive approaches that utilize machine learning (ML) and deep learning (DL) are becoming increasingly popular for network optimization in WiGig networks. However, existing ML/DL models are often tailored to specific network environments, making them susceptible to concept drift — a phenomenon where even minor environmental changes can significantly degrade network performance due to incorrect decision-making. This paper investigates scenarios and environmental changes that can trigger concept drift in WiGig networks. We conduct real-world experiments to analyze the statistical behavior of received signal strength, highlighting the potential for concept drift. Based on our findings, we propose a direction for identifying concept drift in WiGig networks.

The escalating integration of deep neural networks (DNNs) in autonomous vehicles underscores the urgency of fortifying them against adversarial attacks. This paper presents a novel approach to enhance the robustness of convolutional neural networks (CNNs) in self-driving cars through a combination of adversarial mitigation techniques: they are randomization, image padding, and, most uniquely, the addition of random Gaussian noise after convolution layers. Our specialized neural network demonstrates consistent steering control under various attack scenarios, avoiding the over-steering or under-steering issues observed in standard models. As 6 G networks emerge with their ultra-reliable low-latency communication capabilities, our research contributes to the security foundation necessary for autonomous vehicles in this coming era, where resilience against adversarial manipulation will be crucial for maintaining safety in increasingly connected transportation ecosystems. Our open-sourced model provides a benchmark for real-time attackresistant systems applicable to 6G-enabled autonomous driving technologies.

Autonomous vehicles generate massive sensor data daily but operate as isolated intelligence units due to privacy constraints and network limitations. Current centralized machine learning approaches face critical barriers including compliance issues, high bandwidth costs, and latency constraints preventing real-time safety decisions. While Federated Learning (FL) enables collaborative training without raw data sharing and 6G networks promise ultra-low latency, a fundamental mismatch exists between FL's dynamic computational demands and 6G's static resource allocation mechanisms. This paper presents Predictive FL-Aware Network Slicing (PFANS), a novel framework that integrates real-time convergence modeling with proactive 6 G slice reconfiguration for autonomous vehicle networks. PFANS predicts FL computational demands multiple training rounds in advance and automatically reconfigures network slices before bottlenecks occur. Experimental results demonstrate superior resource utilization efficiency, significantly faster convergence compared to baseline approaches, and excellent handover success rates with minimal context migration times. The framework achieves state-of-theart prediction accuracy while introducing negligible network overhead, establishing effective adaptive resource management for next-generation vehicular networks.

The integrity of Internet of Things (IoT) devices has caused a fast spread in an era of data-driven decision-making across businesses. This tutorial survey provides a comprehensive review of current IoT data handling advances, focusing on data quality management (DQM). The article starts with the key aspects of IoT data management. In this regard, we shed light on the data source, volume and velocity, variety, lifecycle, security and privacy, scalability and distribution processing, anomaly detection, and energy efficiency. Then, We present a comprehensive taxonomy of IoT DQM based on the application type, such as smart cities, healthcare, agriculture, environmental monitoring, retail and supply chain, and smart grids (SGs). As IoT data processing, analysis, and security play a significant role in DQM; this tutorial survey carefully addresses how modern technologies maintain this role. More particularly, this work investigates the use of edge computing for real-time data processing and the incorporation of synthetic data to supplement restricted resources incorporating the issues of managing the massive datasets created by IoT implementations. In addition, the paper addresses the use of machine learning (ML) algorithms for in-depth analysis of IoT data streams, DQ evaluation protocols, and detection tactics during data transfer. Moreover, the article investigates complete security measures for protecting sensitive data, such as access control regulations and several security techniques, including authentication, encryption, and secure communication protocols that enable IoT data management. Besides, blockchain technology’s significant roles in this regard have been comprehensively addressed. Along with summarizing and reviewing the latest efforts in DQM in IoT-based systems, we shed light on their strong and weak points and discuss upcoming trends and potential difficulties in IoT data management. Last but not least, we continue by emphasizing the cumulative impact of these advances and shedding light on the open issues and challenges. Finally, this in-depth tutorial survey aims to be a significant resource for academics, practitioners, and stakeholders interested in the changing environment of IoT data management, with a particular emphasis on DQ.

Integrating the Internet of Medical Things (IoMT) and artificial intelligence (AI) is revolutionizing healthcare by enabling real-time health monitoring, predictive analytics, and personalized treatment. However, existing AI healthcare models are trained offline on static datasets, making them less adaptable to evolving health data and potentially reducing their accuracy and decision-making. Furthermore, adversaries may exploit this by injecting frequent data shifts, straining healthcare resources. Privacy concerns also arise from the exposure of sensitive patient data. Therefore, we propose a novel AI-driven healthcare methodology with secure, decentralized, and privacy-enhancing adaptive intelligence. First, a deep learning (DL) model is devised to leverage its high-confidence probability to detect data drift efficiently. Next, we propose a privacy-preserving approach leveraging functional encryption to ensure patient data confidentiality during drift detection and model retraining while eliminating reliance on a trusted entity. Lastly, we propose a customized consortium blockchain with group signatures to protect patient anonymity and data tampering and unlinkability while preventing falsely claiming drift incidents. Moreover, to ensure decentralization, it removes the need for a trusted authority in cryptographic key generation. Our experiments, on a real testbed and healthcare datasets, show that the proposed methodology achieves real-time drift detection with performance comparable to existing methods, while reducing the computational time by 52.35%. It also maintains high accuracy, achieving up to 98.43% with the offline health monitoring model and up to 96% with the online adaptive model. Additionally, it preserves patient privacy while reducing computational and communication overhead by 94.26% and 89%, respectively, compared to the state-of-the-art.

Unmanned Aerial Vehicles (UAVs) play a crucial role in various applications, including detecting environmental hazards, e.g., wildfire smoke detection. However, the limited computational capabilities and battery life of UAVs present barriers to deploying complex artificial intelligence (AI) models onboard. To address this challenge, we propose a novel hybrid deep learning framework for UAVs to carry out light-weight yet efficient smoke detection. The framework combines a lightweight model for initial image assessment and a depth-wise model for selective processing of uncertain cases. Bayesian optimization is employed to determine the optimal threshold values for activating the depth-wise model, striking a balance between accuracy and computational efficiency. The proposed approach eliminates the need for cloud server connectivity, enabling onboard decision-making. Experimental results demonstrate that the hybrid framework achieves significant reductions in processing time and the number of calls to the depth-wise model while maintaining high accuracy. The framework’s adaptability and robustness make it suitable for real-time smoke detection applications in resource-constrained environments.

Cell-free networks have emerged as a new paradigm for beyond-5G networks, offering uniform coverage and improved control over interference. However, scalability poses a challenge in full cell-free networks, where all access points (APs) serve all users. This challenge is addressed by user-centric clustering, where each user is served by a subset of APs, reducing complexity while maintaining coverage. In this paper, we provide an analysis of the relation between the user-centric clustering and pilot assignment problems in cell-free networks, and introduce a formulation which decouples both problems enabling each to be solved independently. We present a general problem formulation for the user-centric clustering problem, allowing the use of diverse per-user and network-wide performance metrics. Specifically, we focus on one instance of this framework, utilizing per-user spectral efficiency and network-wide sum spectral efficiency (SE) as metrics. Additionally, we formulate the pilot assignment problem to minimize overall channel estimation error while considering the user-centric clusters in evaluating the desirability of pilot assignments, which leads to better performing solutions. Both problems are classified as binary nonlinear programs that are at least NP-hard. To solve these optimization problems, our proposed methodology employs sample average approximation coupled with surrogate optimization for the user-centric clustering problem and utilizes the genetic algorithm for the pilot assignment problem. Numerical experiments demonstrate that the optimized solutions surpass baseline solutions, leading to significant improvements in spectral efficiency.

The convergence of Intelligent Reflecting Surfaces (IRS) and Terahertz (THz) communications represents a transformative advancement for sixth-generation (6G) wireless networks, yet presents unprecedented challenges in system optimization. This paper addresses the critical challenge of joint optimization between IRS phase shifts and THz resource allocation in dynamic Internet of Things (IoT) environments, focusing on real-time adaptation to rapidly changing channel conditions. We propose a novel Adaptive Online Multi-Agent Deep Deterministic Policy Gradient (MADDPG) framework that leverages dynamic experience weighting to automatically adjust learning based on detected environmental changes. Our approach incorporates a multi-resolution buffer structure that balances recent observations with historical patterns, enabling both rapid adaptation and long-term optimization while considering the unique characteristics of THz-band propagation and IRS reflection patterns. The framework employs explicit coordination protocols between IRS controllers and resource managers, significantly improving convergence in non-stationary environments. Comprehensive simulations using realistic THz channel models and practical IRS configurations demonstrate that our proposed framework achieves a 45% improvement in system throughput, a 38% reduction in end-to-end latency, and a 30% enhancement in energy efficiency compared to conventional optimization approaches. More significantly, our solution demonstrates unprecedented adaptation capabilities, recovering 90% of optimal performance within 5 ms after abrupt environmental changes a critical requirement for future 6G networks. The framework maintains robust performance under diverse conditions, including high user mobility scenarios and adverse atmospheric conditions, while exhibiting linear computational scaling with increasing IRS elements (tested up to 512 elements). These results establish the viability of Adaptive Online MADDPG-based joint IRS-THz optimization for practical 6G deployments, particularly in dynamic IoT environments where traditional communication approaches face significant limitations.

Existing literature confirms the ability of machine learning to identify fraudulent smart grid power consumers who report false consumption readings to pay less electricity bills. Additionally, federated learning (FL) shows promise as a way to train the detection model without requiring data sharing, thereby safeguarding consumer privacy. However, malicious participants (i.e., clients) in FL training can launch adversarial attacks by training their local models with specially crafted low-consumption data to inject a Trojan into the global model. This Trojan can then be activated during the evaluation phase to evade the detection of false data. To the best of our knowledge, not enough research has been done on this topic in the context of unsupervised learning. The absence of labels in unsupervised learning exacerbates the effectiveness of Trojan attacks and renders it more challenging to design robust defense mechanisms. In this article, we first investigate the vulnerability of one-class classifiers to Trojan attacks. Then, we propose two defense approaches named layerwise close-to-median (LWCM) and Machine Unlearning to counter this attack. In LWCM, by choosing a FL client whose last layer model parameters are near to the median of all clients’ last layer parameters to update the global model, we can identify and exclude malicious updates. The idea is that the last layer parameters of honest clients should be similar, whereas those from malicious clients are different. With the majority of clients being honest, the median values are closer to the parameters of these clients, facilitating the detection of malicious clients. In Machine Unlearning, we utilize gradient ascent-based techniques to adapt models by selectively removing attacker-related data points. This is possible because honest clients generate data resembling that of malicious clients and employ a dual-component loss function to maintain model proficiency in recognizing benign power consumption patterns while eliminating malicious patterns. To show the seriousness of Trojan attacks and the effectiveness of our countermeasures, many experiments have been carried out.

In the advanced metering infrastructure (AMI) of the smart grid, smart meters (SMs) are deployed to collect fine-grained electricity consumption data, enabling billing, load monitoring, and efficient energy management. However, some consumers engage in fraudulent behavior by hacking their meters, leading to either traditional electricity theft or more sophisticated evasion attacks. Evasion attacks aim to illegally reduce electricity bills while deceiving theft detection mechanisms. The current methods for identifying such attacks raise privacy concerns due to the need for access to consumers' detailed consumption data to train detection mechanisms. To address privacy concerns, federated learning (FL) is proposed as a collaborative training approach across multiple consumers. Adversarial training (AT) has shown promise in countering evasion threats on machine learning models. This paper, first, investigates the susceptibility of traditional electricity theft classifiers trained by FL to evasion attacks for both independent and identically distributed (IID) and Non-IID consumption data. Then, it investigates the effectiveness of AT in securing the global electricity theft detector against evasion attacks, assuming no misbehavior from the participant consumers in the FL process. After that, we introduce a novel attack, called Distillation, which can be launched during the AT process to make the global model susceptible to evasion at inference time. Finally, extensive experiments are conducted to validate the severity of the proposed attack.

In smart power grids, certain customers are motivated by financial gains to manipulate electricity consumption data, aiming to reduce their bills. Despite the development of machine learning-based detectors, these systems remain vulnerable to evasion attacks. This paper investigates the susceptibility of deep reinforcement learning (DRL)-based detectors to evasion attacks. We propose an evasion attack model that employs the double deep Q learning (DDQN) algorithm for a black-box attack scenario. Our model generates adversarial evasion samples by altering malicious consumption data, tricking detectors into classifying them as benign. Leveraging the unique attributes of reinforcement learning (RL), our model determines optimal actions for manipulating malicious data. For comparative analysis, we compare our DRL-based model with an FGSM-based attack model. Our experiments consistently demonstrate the effectiveness of our DRL-based attack model, achieving an impressive attack success rate (ASR) ranging from 92.92% to 99.96%, outperforming the FGSM-based attack model.

In smart power grids, electricity theft causes huge economic losses to electrical utility companies. Machine learning (ML), especially deep neural network (DNN) models hold state-of-the-art performance in detecting electricity theft cyberattacks. However, DNN models are vulnerable to adversarial attacks, i.e., evasion attacks. In this work, we study the vulnerability of the DNN-based electricity theft detectors against evasion attacks and the influence of the model's regularization (generalization) on robustness. We cluster the power consumers and train a detector for each cluster, and compare the performance and robustness of this detector to a global detector that is trained on all the consumers, data. The results indicate that the cluster-based detector is not only more robust against evasion attacks but also enhances normal classification accuracy because its training data has more consumption pattern similarity compared to the training data of the global detector which requires higher level of regularization. Moreover, unlike the existing solutions that sacrifice the normal accuracy of the model to improve the robustness against evasion attacks, the proposed cluster-based detector holds state-of-the-art performance in both robustness and accuracy.

Cloud-based federated learning (FL) services have received increasing attention due to their ability to enable collaborative global model training without the need to collect local data from participants. To generate a global model, local models are trained on participants' local data and only model parameters are sent to an aggregator server. Nonetheless, revealing model parameters can still reveal training data via launching attacks, e.g., inference and membership. Hence, to protect model parameters, a secure global model aggregation scheme is needed to protect these parameters from unauthorized access. Existing solutions to this issue, which are based on homomorphic encryption and secure multi-party computation, tend to have large overheads and slow down training times. Functional encryption (FE) has been proposed as a solution for resolving privacy-preservation issues in FL, but current solutions suffer from high overhead and lack of security such as leaking master private key. To address these issues, this paper proposes a privacy-protecting, efficient, and decentralized FL framework, called FedSafe, based on FE without the need for a trusted key distribution center (KDC). The proposed scheme allows the participants to communicate with an aggregator to construct a global model without disclosing or learning their local models' parameters or the training data, thereby safeguarding their privacy. Through rigorous testing with real-world data, it is demonstrated that FedSafe outperforms the state-of-the-art privacy-protecting FL schemes in terms of security, scalability, and communication and computation overhead. Unlike existing approaches, this is accomplished without depending on any trusted KDC.

This paper explores secret key generation in 5G and beyond LiFi networks using visible light in the downlink and infrared in the uplink. Unlike the existing works, we focus on a realistic indoor environment with multi-user mobility. Given inaccuracies in high-frequency channel models, we introduce the first deep learning model that combines the channel probing and quantization phases to generate initial secret keys with a minimal key disagreement rate (KDR) of 16% between the uplink and downlink, leading to a key generation rate (KGR) of 79 bits/s after information reconciliation. We show that LiFi channel statistics suffer from concept drifts with user density changes in the room. This increases the KDR by 28% - 44% and the generated keys fail to pass the NIST randomness tests. As a countermeasure, we introduce a voting ensemble model that mitigates concept drifts, maintaining a stable 16% KDR, 79 bits/s KGR, and passing NIST tests, despite the varying user densities.

The utilization of Advanced Metering Infrastructure (AMI) technology is for recording and billing customers for electricity consumption. This technology is vulnerable to cyber-attacks where customers under report their electricity usage, causing financial losses for electricity providers. Machine learning (ML) can be used to detect electricity theft, but it is challenging due to the absence of malicious data. To address this challenge, most of the existing works proposed specific attacks, however, these works are only effective on the proposed attacks and fail on new attacks. Some works proposed using anomaly detectors trained only on begin dataset. However, they rely on specific attacks to set classification thresholds, leading to failure in detecting zero-day attacks. Therefore, this paper proposes a one-class classification approach for electricity theft detection depending only on benign data and without assuming any attacks. First, the paper re-evaluates an existing detector that sets a reconstruction error threshold. Then, it proposes a detector combining decisions from three one-class ML models, including a one-class support vector machine (OC-SVM) trained on benign data, an OC-SVM trained on the bottleneck outputs of an autoencoder trained only on benign data, an OC-SVM trained on the mean squared errors of the reconstructed data of the autoencoder. The evaluation results confirm the superiority of the proposed detector over its individual components and the existing detectors.

Artificial Intelligence (AI) applications on Internet of Things (IoT) networks often involve relaying generated data to a server for deep learning training, which poses security risks to users' data. Federated Learning (FL) offers a distributed model training paradigm in which local data are kept at the edge and locally trained models are exchanged and aggregated by a server over several rounds to produce a global model. While successful, standard FL algorithms do not support heterogeneous local model design, an essential requirement, especially for resource-limited edge devices. Recently, Knowledge Distillation-based FL algorithms have provided model-agnostic FL to enable clients to independently design their local model and share soft labels instead of model parameters. KD-based FL algorithms are computationally expensive due to additional distillation training. We propose Federated Learning with Selective Knowledge Distillation (FedSKD) to address the limitations of system heterogeneity; and computation and communication demands. We evaluate different aspects of the proposed algorithm relative to baseline FL algorithms. Results show that FedSKD incurs significantly less per-round computation time and communication overhead relative to the considered model-based and KD-based FL algorithms.

Load forecasting (LF) in smart grids is beneficial not only in mitigating equipment failures and power outages but also in facilitating effective power dispatching and infrastructure planning. To predict future loads accurately, the consumers' fine-grained energy consumption readings are fed into machine-learning (ML) models. However, revealing these readings enables adversaries to deduce confidential information about consumers, including details about their lifestyle, and hence their privacy is violated. To address this privacy issue, the existing works only focus on using federated learning (FL)-based approaches to train and obtain an accurate global LF model. Nevertheless, addressing the privacy violation problem during the LF process (in the deployment phase) after obtaining the global model for AMI networks has not been well investigated yet. Therefore, this paper proposes a novel, efficient, and decentralized approach that enhances the precision of LF while safeguarding the privacy of consumers. The proposed scheme incorporates inner product functional encryption (IPFE) to allow smart meters (SMs) to encrypt their readings with no need for a trusted key distribution center (KDC) while allowing LF without divulging or acquiring knowledge of the consumers' readings to protect their privacy. In addition, a hybrid deep learning approach is developed to construct an LF model that can yield precise forecasts. To show the feasibility of the proposed scheme, the performance of our scheme was assessed on a real energy consumption readings dataset, and the results demonstrate proficiency in LF while providing robustness and privacy preservation with reasonable communication efficiency.

Accurate channel prediction using deep learning (DL) algorithms can address the challenges of terahertz (THz) propagation, such as atmospheric absorption and object scattering, by enabling proactive handover and beamforming. However, indoor environments are inherently dynamic, with factors like occupancy level variations causing the channel characteristics to change over time. This phenomenon, known as concept drift, can severely degrade the DL model performance used in channel prediction. This paper investigates the impact of indoor occupancy level variations on the generalization ability of state-of-the-art DL models for THz channel prediction. We identify three distinct occupancy levels (low, medium, and high) within the THz indoor channel. Our results demonstrate that the state-of-the-art DL models exhibit limited generalization capabilities, with performance deterioration in prediction accuracy ranging from 4−62%. We propose a robust two-stage framework to mitigate concept drift in THz channel prediction. The first stage predicts the indoor occupancy level from the THz wireless signal, which is a multi-class classification problem. Due to the reoccurring concept of occupancy levels, the second stage contains a pool of models in a sleeping mode based on a hybrid convolutional neural network (CNN) long-short-term memory (LSTM) architecture. One of these DL expert models is activated for channel prediction based on the occupancy level predicted from the previous stage. Our framework demonstrates superior generalization by limiting the performance deterioration from 62% due to concept drift to ≤ 9%. This represents an 85% reduction in performance deterioration compared to the existing state-of-the-art DL models.

This paper explores the vulnerability of wireless secret key generation (WSKG) to eavesdropping in a dynamic indoor light-fidelity (LiFi) network. It analyzes the channel impulse response (CIR) similarities of two moving user equipments (UEs) across scenarios with two, four, and eight UEs. We observe that as the number of UEs increases, the similarity in CIR also rises, due to the proximal movement patterns among UEs. Specifically, the similarity rate peaks at 70% when eight UEs enter the room; it then drops to 24% during the wandering phase and rises again to 80% as UEs exit the room. Consequently, an eavesdropper among the eight UEs is able to generate 27% of a legitimate UE’s secret key, it significantly reduces the key’s complexity, decreasing the number of possible keys that need to be tested to break the encryption and making it easier to predict the remainder of the key. To mitigate this issue, we introduce a novel approach that utilizes a generative adversarial network (GAN) to artificially manipulate the CIR, thereby reducing the effectiveness of eavesdropping by adding noise into the observed CIR. This method effectively reduces the CIR similarity to a negligible 1%, thus ensuring the integrity of WSKG against eavesdropping threats.

An innovative method to raise wireless communication systems’ efficiency is to use Reconfigurable Intelligent Surface (RIS). Unfortunately, determining the quantity and locations of the RIS elements continues to be difficult, requiring a clever optimization framework. Concerning the practical overlap between the related multi-RISs in wireless communication systems, this paper attempts to minimize the number of RISs while considering the average possible data rate and technological constraints. In this regard, a novel Enhanced Artificial Rabbits Algorithm (EARA) is developed to minimize the number of RISs to be installed. The novel EARA is inspired by the natural survival strategies of rabbits, including detour eating and random concealment. A more effective method of exploring the search space around the best solution so far is produced by the suggested EARA by combining an upgraded Collaborative Searching Operator (CSO) arrangement. Also, an adaptive time function is included to increase the effect of this exploitation tactic by the increasing number of iterations. The simulation results show that the suggested EARA is highly efficient in reaching the maximum success rate of producing the smallest number of RISs under various feasible rate threshold settings. When EARA is compared to standard Artificial Rabbits Optimizer (ARO), Growth Optimizer (GO), Artificial Ecosystem Optimizer (AEO), and Particle Swarm Optimization (PSO), the average number of RISs is improved by 5.32%, 6.7%, 16.73%, and 20.06%, respectively. Furthermore, according to simulation data, the EARA outperforms AEO, GO, ARO, and PSO in terms of success rate at δ=1.4 by 6.66%, 6.66%, 45.43%, and 99%, respectively.

The adoption of Internet of Things (IoT) and monitoring devices in 5G and beyond networks has been widespread. Unmanned aerial vehicles (UAVs) have shown success in connecting rural and remote areas due to the high cost of deploying infrastructures like cellular network base stations and optical fiber connections in vast landscapes with sparse populations. The constrained energy of UAVs results in limited coverage area and flight time, which in turn reduces the potential of UAVs to provide task-oriented wireless communication links. In this article, we explore path optimization and transmission organization algorithms to minimize flight time and extend the range of UAVs performing collaborative federated learning (FL) among geographically dispersed nodes communicating through wireless connections offered by UAVs coupled with device-to-device (D2D) networks. The UAV orchestrates FL between spatially scattered homes via long-range radio wireless communication. We formulate the drone path optimization as a traveling salesman problem (TSP) and employ self-organizing maps (SOM) for path planning. Additionally, knowledge distillation (KD)-based FL is used to reduce communication overhead for the resource-constrained UAV-IoT system. Experimental results demonstrate SOM’s ability to represent the topological structure of nodes and produce a cost-efficient Hamiltonian cycle, from which the drone path is derived. Our results demonstrate the communication efficiency and utility of KD-based FL compared to model-based FL methods. The proposed hybrid solution enables energy-constrained UAVs to perform FL over large areas leveraging a shared data set for KD and a SOM-based path optimization algorithm.

Load forecasting (LF) is a crucial process of predicting future energy load and demand in smart grids, allowing for mitigating equipment failures and power outages, besides facilitating effective power dispatching and infrastructure planning. Methods used in LF range from traditional statistical and mathematical models to modern machine learning (ML) algorithms, and it has been proven that the latter has a better ability to predict future loads. These techniques leverage the consumers’ energy consumption readings for use in the LF process; however, revealing these readings exposes sensitive consumer lifestyle information, hence violating their privacy. The majority of the existing works focus on obtaining precise LF and addressing consumers’ privacy violations during the LF process in the deployment phase has not been well-investigated yet. Moreover, the existing methods that can be employed to preserve privacy introduce high overhead and rely on a trusted third party, which undermines the trust assumption, making them less robust. Therefore, this article proposes a novel and efficient privacy-preserving LF scheme, called privacy-preserving LF and monitoring and billing (PLFMB), that utilizes inner product functional encryption (IPFE) and eliminates the need for a trusted key distribution center. PLFMB allows smart meters to encrypt their consumption readings while enabling the system operator (SO) to 1) evaluate a hybrid deep learning-based LF model developed to predict future loads accurately; 2) monitor the grid load; and 3) compute consumers’ bills following dynamic pricing, without revealing or learning consumers’ readings to protect their privacy. Our proposed scheme has been evaluated on a real energy consumption data set, demonstrating its feasibility, proficiency in LF, and robustness in preserving consumer’s privacy while maintaining reasonable overhead.

Inherent bias in the artificial intelligence (AI)-model brings inaccuracies and variabilities during clinical deployment of the model. It is challenging to recognize the source of bias in AI-model due to variations in datasets and black box nature of system design. Additionally, there is no distinct process to identify the potential source of bias in the AI-model. To the best of our knowledge, this is the first review of its kind that addresses the bias in AI-model by categorizing 48 studies into three classes, namely, point-based, image-based, and hybrid-based AI-models. Selection strategy using PRISMA is adopted to select the 72 crucial AI studies for identifying bias in AI models. Using the three classes, bias is identified in these studies based on 44 critical AI attributes. Bias in the AI-models is computed by analytical, butterfly, and ranking-based bias models. These bias models were evaluated using two experts and compared using variability analysis. AI-studies that lacked sufficient AI-attributes are more prone to risk-of-bias (RoB) in all three classes. Studies with high RoB loses fins in the butterfly model. It has been analyzed that the majority of the studies in healthcare suffer from data bias and algorithmic bias due to incomplete specifications mentioned in the design protocol and weak AI design exploited for prediction.

Earthquake early warning systems (EEWSs) often rely on fast determination of earthquake source parameters, namely, location, magnitude, and depth. In areas where the seismic network is coarse, the capability to determine source parameters based on data recorded by a single station is desirable. Moreover, being able to use a single component of the seismic data might increase the robustness of the system to sensor malfunction and might save on sensor cost and computation time. Here, we propose a hybrid deep learning (DL) model to estimate source parameters based on single-component data recorded by a single station at 3 s after the P-wave onset. The model, which we call EEWS-311, uses a convolutional neural network (CNN) and bidirectional long short-term memory. It is trained and tested on recordings of more than 14000 events by a single station of the Japanese Hi-net high-sensitivity short-period seismic network. Compared with source parameters obtained by conventional methods, our model achieves excellent performance (average errors in latitude, longitude, magnitude, and depth equal to 0.05°, 0.1°, 0.14 velocity magnitude (Mv), and 5.68 km, respectively). The results demonstrate the suitability of EEWS-311 for earthquake early warning in areas with sufficient training data.

Despite the extensive use of IoT and mobile devices in the different applications, their computing power, memory, and battery life are still limited. Multi-Access Edge Computing (MEC) has recently emerged to address the drawbacks of these limitations. With MEC on the network's edge, mobile and IoT devices can offload their computing operations to adjacent edge servers or remote cloud servers. However, task offloading is still a challenging research issue, and it is necessary to improve the overall Quality of Service (QoS) and attain optimized performance and resource utilization. Another crucial issue that is usually overlooked while handling this matter is offloading an application that consists of dependent tasks. In this study, we suggest a Refined Whale Optimization Algorithm (RWOA) for solving the multiuser dependent tasks offloading problem in the Edge-Cloud computing environment with three objectives: 1- minimizing the application execution latency, 2- minimizing the energy consumption of end devices, and 3- the charging cost for used resources. We also avoid the traditional binary planning mechanisms by allowing each task to be partially processed simultaneously at three processing locations (local device, MEC, cloud). We compare RWOA with other Optimizers, and the results demonstrate that the RWOA has optimized the fitness by 52.7% relative to the second best comparison optimizer.

This paper addresses the user-centric clustering and pilot assignment problems in cell-free networks, recognizing the need to solve both problems simultaneously. The motivation of this research stems from the absence of benchmarks, general formulations, and the reliance on subjectively designed objective functions and heuristic algorithms prevalent in existing literature. To tackle these challenges, we formulate stochastic non-linear binary integer programs for both the user-centric clustering and pilot assignment problems. We specifically design the pilot assignment formulation to incorporate user-centric clusters when evaluating the desirability of pilot assignments, resulting in improved efficiency. To solve the problems, the proposed methodology employs sample average approximation coupled with surrogate optimization for the user-centric clustering problem and the genetic algorithm for the pilot assignment problem. Numerical experiments demonstrate that the optimized solutions outperform baseline solutions, leading to significant gains in spectral efficiency.

Federated learning is a promising approach for training models on distributed data, driven by increasing demand in various industries. However, federated learning framework faces several key challenges, including communication bottlenecks and client data heterogeneity. Personalized asynchronous federated learning addresses these challenges by customizing the model for individual users based on their local data while trading model updates asynchronously. In this paper, we propose the Personal-ized Moreau Envelopes-based Asynchronous Federated Learning (APFedMe). Our approach combines the strengths of Moreau En-velopes to handle optimization problems and asynchronous weight updates to improve communication efficiency while mitigating heterogeneity data challenges through a personalized learning environment. We evaluate our approach on several datasets and compare it with the baseline PFedMe method. Our experiments demonstrate that the proposed APFedMe outperforms other meth-ods in terms of convergence speed and communication efficiency. Overall, our work contributes to developing more effective and efficient federated learning methods that can be applied in various real-world scenarios.

Medical image segmentation is a crucial component of computer-aided diagnosis (CAD) systems, as it aids in identifying important areas in medical images. In order to achieve optimal segmentation results, it is important to preserve the resolution of the input image. The dilated convolution module was introduced to maintain resolution across layers of a deep convolutional neural network by increasing the receptive field exponentially while keeping the parameters increase linearly. However, one drawback of using dilated convolution is that it can result in local spatial resolution loss by increasing the sparsity of the kernel in checkboard patterns. This work proposes a double-dilated convolution module to maintain local spatial resolution in medical image segmentation tasks while having a large receptive field. The module is applied to tumor segmentation in breast cancer mammograms using the state-of-art Deeplabv3+ network. The study also evaluates the developed models with the Gradient weighted Class Activation Map (Grad-CAM) and compares the performance of lesion segmentation networks on mammogram screenings from the INBreast dataset before and after using the proposed dilation module. The results show that the proposed module effectively improves the segmentation performance.

Federated Learning (FL) has emerged as a powerful approach to facilitate the construction of centralized models without compromising the data privacy of multiple participants. However, conventional FL methodologies do not address system heterogeneity where each participant needs to independently design its own model, a prevalent requirement in Internet of Things (IoT) applications due to the heterogeneous nature of tasks and data. Knowledge Distillation-based FL algorithms tackle this limitation by exchanging soft labels instead of model weights, thus giving each client the ability to independently design its local model architecture. While FL is inherently private, studies have indicated that exploiting gradients for a few iterations can reveal sensitive training data. To protect against privacy attacks, FL algorithms employ Differential Privacy (DP) to guarantee privacy protection, which can be applied using Local Differential Privacy (LDP). In this paper, we elaborate on preserving clients' training data privacy in KD (Knowledge Distillation)-based FL using DP, providing both privacy and flexibility. We provide theoretical analysis to extend the privacy guarantee to exchanged updates. Experimental analysis is performed utilizing Human Activity Recognition (HAR) datasets with different modalities. The results obtained demonstrate the capacity of KD-based FL to maintain a robust utility-privacy balance. Furthermore, for the same DP protection level, the utility of models trained on images was severely reduced across all FL algorithms. This suggests that the modality and complexity of a dataset are important factors for shaping the utility-privacy tradeoff of DP.

As vehicular communication networks embrace metaverse beyond 5G/6G systems, the rich content update via the least interfered subchannel of the optimal frequency band in a hybrid band vehicle to everything (V2X) setting emerges as a challenging optimization problem. We model this problem as a tradeoff between multi-band VR/AR devices attempting to perform metaverse scenes and environmental updates to metaverse roadside units (MRSUs) while minimizing energy consumption. Due to the computational hardness of this optimization, we formulate an opportunistic band selection problem using a multi-armed bandit (MAB) that provides a good quality solution in real-time without computationally burdening the already stretched augmented/virtual reality (AR/VR) units acting as transmitting nodes. The opportunistic use of scheduling rich content updates at traffic signals and stand-still scenarios maps well with the formulated bandit problem. We propose a Dual-Objective Minimax Optimal Stochastic Strategy (DOMOSS) as a natural solution to this problem. Through extensive computer-based simulations, we demonstrate the effectiveness of our proposal in contrast to baselines and comparable solutions. We also verify the quality of our solution and the convergence of the proposed strategy.

Emerging Internet of Things (IoT) applications, such as sensor-based Human Activity Recognition (HAR) systems, require efficient machine learning solutions due to their resource-constrained nature which raises the need to design heterogeneous model architectures. Federated Learning (FL) has been used to train distributed deep learning models. However, standard federated learning (fedAvg) does not allow the training of heterogeneous models. Our work addresses the model and statistical heterogeneities of distributed HAR systems. We propose a Federated Learning via Augmented Knowledge Distillation (FedAKD) algorithm for heterogeneous HAR systems and evaluate it on a self-collected sensor-based HAR dataset. Then, Kullback-Leibler (KL) divergence loss is compared with Mean Squared Error (MSE) loss for the Knowledge Distillation (KD) mechanism. Our experiments demonstrate that MSE contributes to a better KD loss than KL. Experiments show that FedAKD is communication-efficient compared with model-dependent FL algorithms and outperforms other KD-based FL methods under the i.i.d. and non-i.i.d. scenarios.

Fifth Generation (5G) networks operating on mmWave frequency bands are anticipated to provide an ultrahigh capacity with low latency to serve mobile users requiring high-end cellular services and emerging metaverse applications. Managing and coordinating the high data rate and throughput among the mmWave 5G Base Stations (BSs) is a challenging task, and it requires intelligent network traffic analysis. While BSs coordination has been traditionally treated as a centralized task, this involves higher latency that may adversely impact the user’s Quality of Service (QoS). In this paper, we address this issue by considering the need for distributed coordination among BSs to maximize spectral efficiency and improve the data rate provided to their users via embedded AI. We present Peer-Coordinated Sequential Split Learning dubbed PC-SSL, which is a distributed learning approach whereby multiple 5G BSs collaborate to train and update deep learning models without disclosing their associated mobile users data, i.e., without privacy leakage. Our proposed PC-SSL minimizes the data transmitted between the client BSs and a server by processing data locally on the clients. This results in low latency and computation overhead in making handoff decisions and other networking operations. We evaluate the performance of our proposed PC-SSL in the mmWave 5G throughput prediction use-case based on a real dataset. The results demonstrate that our proposal outperforms conventional approaches and achieves a comparable performance to centralized, vanilla split learning.

As 5G and beyond networks continue to proliferate, intelligent monitoring systems are becoming increasingly prevalent. However, geographically isolated regions with sparse populations still face difficulties in accessing these technologies due to infrastructure deployment challenges. Additionally, the high cost and unreliability of satellite Internet services make them less appealing. This paper studies the challenges of drone-aided networks and presents a communication-efficient Federated Learning (FL) system on a drone-aided Internet of Things (IoT) network to facilitate health analysis in rural areas over LoRa wireless links. The proposed approach consists of two primary components. Firstly, optimizing the drone’s trajectory is theoretically formulated as a modified version of the Traveling Salesman Problem (TSP), with the Self-Organizing Map (SOM) algorithm employed for effective route planning. Secondly, the Knowledge Distillation (KD)-based FL algorithm is utilized to reduce communication overhead by leveraging soft labels. The quality of drone routes generated by the SOM is evaluated on multi-scale maps with pre-determined optimal paths. The experiments reveal SOM’s ability to accurately represent node topologies and yield cost-effective Hamiltonian cycles. The KD-based FL proves to be more efficient in terms of communication than FedAvg as the former exchanges soft labels while the latter exchanges model weights, thus reducing drone waiting time and battery consumption. We showcase the performance of our KD-based FL algorithm using Human Activity Recognition (HAR) datasets, illustrating a communication-efficient alternative for distributed learning, offering competitive performance leveraging a shared dataset for knowledge transfer among IoT devices.

Recently emerging WiGig systems experience limited coverage and signal strength fluctuations due to strict line-of-sight (LoS) connectivity requirements. In this paper, we address these shortcomings of WiGig communication by exploiting two emerging technologies in tandem, namely the reconfigurable intelligent surface (RIS) and unmanned aerial vehicles (UAVs). In ultra-dense traffic sites (referred to as hotspots) where WiGig nodes or User Devices (UDs) experience complex propagation and non-line-of-sight (non-LoS) environment, we envision the deployment of a UAV-mounted RIS system to complement the WiGig base station (WGBS) to deliver services to the UDs. However, commercially available UAVs have limited energy (i.e., constrained flight time). Therefore, the trajectory of our considered UAV needs to be locally estimated to enable it to serve multiple hotspots while minimizing its energy consumption within the WGBS coverage boundaries. Since this tradeoff problem is computationally expensive for the resource-constrained UAV, we argue that sequential learning can be a lightweight yet effective solution to locally solve the problem with a low impact on the available energy on the UAV. We formally formulate this problem as a contextual multi-armed bandit (CMAB) game. Then, we develop the linear randomized upper confidence bound (Lin-RUCB) algorithm to solve the problem effectively. We regard the UAV as the bandit learner, which attempts to maximize its attainable rate (i.e., the reward) by serving distinct hotspots in its trajectory that we treat as the arms of the considered bandit. The context is defined as the hotspots’ locations provided using GPS (global positioning system) service and the reward history of each hotspot. Our proposal accounts for the energy expenditure of the UAV in moving from one hotspot to another within its battery charge lifetime. We evaluate the performance of our proposal via extensive simulations that exhibit the superiority of our proposed Lin-RUCB algorithm over benchmarking methods.

The mmWave WiGig frequency band can support high throughput and low latency emerging applications. In this context, accurate prediction of channel gain enables seamless connectivity with user mobility via proactive handover and beamforming. Machine learning techniques have been widely adopted in literature for mmWave channel prediction. However, the existing techniques assume that the indoor mmWave channel follows a stationary stochastic process. This paper demonstrates that indoor WiGig mmWave channels are non-stationary where the channel’s cumulative distribution function (CDF) changes with the user’s spatio-temporal mobility. Specifically, we show significant differences in the empirical CDF of the channel gain based on the user’s mobility stage, namely, room entering, wandering, and exiting. Thus, the dynamic WiGig mmWave indoor channel suffers from concept drift that impedes the generalization ability of deep learning-based channel prediction models. Our results demonstrate that a state-of-the-art deep learning channel prediction model based on a hybrid convolutional neural network (CNN) long-short-term memory (LSTM) recurrent neural network suffers from a deterioration in the prediction accuracy by 11–68% depending on the user’s mobility stage and the model’s training. To mitigate the negative effect of concept drift and improve the generalization ability of the channel prediction model, we develop a robust deep learning model based on an ensemble strategy. Our results show that the weight average ensemble-based model maintains a stable prediction that keeps the performance deterioration below 4%.

Pelvis fracture detection is vital for diagnosing patients and making treatment decisions for traumatic pelvis injuries. Computer‐aided diagnostic approaches have recently become popular for assisting doctors in disease diagnosis, making their conclusions more trustworthy and error‐free. Inspecting X‐ray images with fractures needs a lot of time from experienced physicians. However, there is a lack of inexperienced radiologists in many hospitals to deal with these images. Therefore, this study presents an accurate computer‐aided‐diagnosing system based on deep learning for detecting pelvis fractures. In this research, we construct an explainable artificial intelligence (XAI) framework for pelvis fracture classification. We used a dataset containing 876 X‐ray images (472 pelvis fractures and 404 normal images) to train the model. The obtained results are 98.5%, 98.5%, 98.5%, and 98.5% for accuracy, sensitivity, specificity, and precision.

Energy theft causes economic losses and power out-ages and disrupts energy generation and distribution of smart grids. A significant challenge is how to effectively use customers' power consumption data for energy theft detection while pre-serving security and privacy. One solution is to use federated learning (FL) to compute a global model to detect energy theft cyberattacks where detection stations train local models on their customers' power consumption data and send only the parameters of the models to an aggregator server. Nevertheless, revealing the model's parameters may still leak customers' private data by launching attacks such as membership and inference. Therefore, a secure aggregation scheme is needed to protect the models' param-eters. Furthermore, the existing privacy-preserving aggregation schemes suffer from high overhead and low model accuracy. This paper addresses these limitations by proposing a novel privacy- preserving, efficient, decentralized, aggregation scheme based on a functional encryption cryptosystem for energy theft detection in smart grids without requiring a key distribution center. Our scheme enables the detection stations to send encrypted training parameters to an aggregator, which calculates the aggregated parameters and returns the updated model parameters to the detection stations without being able to learn the parameters of the local models or the training data of the customers to preserve their privacy. Moreover, the results of our extensive experiments show that our FL-based detector can detect energy thefts accurately with low overhead because of our lightweight privacy-preserving aggregation scheme.

This paper investigates wireless secret key generation in 5G+ indoor LiFi networks operating in the visible light band in the downlink and the infrared band in the uplink. Unlike the existing research, this paper aims to characterize the impact of indoor user mobility on the ability to generate identical secret keys between the user and the base station. Hence, we first generate mobility traces that reflect realistic human behavior in an indoor setup to capture the sensitivity of the optical channels to random blockage and transceiver-misorientation events associated with user mobility. Next, we generate the corresponding channel impulse response and adopt guard band quantization and information reconciliation to extract identical uplink and downlink keys. Then, we examine a set of spatial and temporal features related to the channel probing rate, key disagreement rate, and key generation rate, along with their statistical distributions. Further, we study the randomness of the key following the NIST randomness tests. Our study demonstrated that the handover strategy significantly impacts the randomness of the key and its generation rate.

This paper investigates the performance of non-orthogonal multiple access (NOMA) based hybrid satellite-unmanned aerial vehicle (UAV) systems, where a low Earth orbit (LEO) satellite communicates with the ground users via a decode and forward (DF) UAV relay. We investigate a two NOMA users system, where a far user (FU) and a near user (NU) are served by the UAV which is located at a certain height above the origin of the coverage circle. The channel between satellite and UAV is assumed to follow a Shadowed-Rician fading and the channels between UAV and users are assumed to follow a Nakagami-m fading. New closed-form expressions of the outage probabilities for the two users and the system are derived. Different from other work in literature, we take into consideration different parameters affecting the total link budget. Additionally, we propose an algorithm for minimizing the system outage probability. The mathematical analysis is verified by extensive representative Monte-Carlo (MC) simulations. Finally, simulations are provided to demonstrate the impact of important parameters on the considered system as well as the superiority of the NOMA scheme the over reference scheme.

General Purpose Vision System (GPVS) is a task-agnostic vision-language system that inputs an image and a question from which the system recognizes the tasks to be performed and outputs bounding boxes, confidence scores, and text outputs to answer the question. While much attention to GPVS has been recently given in the computer vision field, its medical field applications are still in their infancy. This paper presents MED-GPVS, a customized deep learning-based GPVS on biomedical images to perform various vision tasks, such as object detection and visual question answering, on medical images to facilitate precision medicine/e-health services. Our envisioned MED-GPVS takes an image and a natural language text as inputs, and then outputs bounding boxes, confidence scores, and generates a caption (i.e., the answer to the posed query). For example, if a medical image of a patient’s abdomen is presented to MED-GPVS followed by the question: "does the picture contain stomach?", MED-GPVS should ideally provide the answer "yes" along with a prediction box and prediction score on the image. We utilize the multilingual SLAKE dataset, which was annotated by expert physicians with a full semantic label, to validate the performance of MED-GPVS under various scenarios involving different biomedical image-based diagnoses. For the visual question answering (VQA) task, MED-GPVS demonstrates encouraging performance with significantly high accuracy of 82.41%.

During the recent years, online users, particularly in social networks, have witnessed an upsurge in racism, sexism, and other types of aggressive and cyberbully content, which are often manifested through offensive, abusive, or hateful speech and harassment. This can lead to severe physical and psychological stress in young children and adolescents, leading to even suicides and negatively affecting social policies. Therefore, there is a significant need to identify and regulate harassing content posted on the Internet in a smart, automated, and accurate manner. With this aim, in this paper, we design and develop a hierarchical framework comprising machine learning algorithms in order of higher computational complexity to adaptatively switch among them for efficiently detecting hateful and abusive content. We combine simple machine learning models such as Naive Bayes/Logistic Regression classifiers with customized calibration and Expectation-Maximization (EM) algorithms, and compare them with the much stronger deep learning techniques. Our proposed hierarchical framework demonstrates a significant improvement of the automated detection of abusive contents in social networks with a relatively small twitter dataset in contrast with the deep learning-based counterpart, namely the Bidirectional Encoder Representations from Transformers (BERT) model, training of which typically requires a much higher volume of labeled documents to detect abusive comments.

The recent COVID-19 (novel coronavirus disease) pandemic induced a deep polarization among regional as well as global communities. The sentiments regarding the pandemic and its impact on lifestyle and economy, often expressed via social networks, are regarded as critical metrics for capturing such polarization and formulating appropriate intervention by the relevant authorities. While there exist a myriad of Natural Language Processing (NLP) models for mining social media data, we demonstrate the shortcomings of the individual models in this paper, and explore how to improve the COVID-19 sentiment analysis in social media network data via two hybrid predictive models based on a Long-Short-Term-Memory (LSTM)-based autoencoder and a Convolutional Neural Network (CNN) model coupled with a bi-directional LSTM. Through extensive experiments on the recently acquired Twitter dataset, we compare the COVID-19 sentiments exhibited in the USA and Canada using our proposed hybrid predictive models and demonstrate their superiority over individual Artificial Intelligence (AI) models.

In the smart grid, malicious customers may compromise their smart meters (SMs) to report false readings to achieve financial gains illegally. This causes hefty financial losses to the utility and may degrade the grid performance because the reported readings are used for energy management. This article is the first work that investigates this problem in the net-metering system, in which one SM is used to report the difference between the power consumed and the power generated. First, we prepare a benign data set for the net-metering system by processing a real power consumption and generation data set. Then, we propose a new set of attacks tailored for the net-metering system to create a malicious data set. After that, we analyzed the data and found time correlations between the net meter readings and correlations between the readings and relevant data obtained from trustworthy sources, such as solar irradiance and temperature. Based on the data analysis, we propose a general multidata-source deep hybrid learning-based detector to identify the false-reading attacks. Our detector is trained on net meter readings of all customers besides data from trustworthy sources to enhance the detector performance by learning the correlations between them. The rationale here is that although an attacker can report false readings, he cannot manipulate the solar irradiance and temperature values because they are beyond his control. Extensive experiments have been conducted, and the results indicate that our detector can identify the false-reading attacks with a high detection rate of 98.59% and a low false alarm of 2.92%.

To handle the exponentially growing service expectations from mobile users and circumvent the band switching slow rate, device-to-device (D2D) communication is receiving much research attention in the Internet of Things (IoT). While the emerging D2D nodes can support heterogeneous frequency bands [radio frequency (RF) including 2.4 GHz/5 GHz wireless local area network (WLAN), 38-GHz millimeter wave (mmWave), and visible light communication (VLC)], the physical constraints (e.g., blocking) require the user devices to dynamically switch between the bands in order to avoid the loss of connectivity and throughput degradation. In this article, we investigate an effective online link selection in hybrid RF-VLC scenarios for direct user data handling. First, we model the multiband selection issue as a multiarmed bandit (MAB) problem. The source/relay node acts as a player who gambles to maximize its long-term feedback/reward via selecting suitable arms, i.e., available bands (WLAN, mmWave, or VLC). Then, we propose an online, energy-aware band selection (EABS) methodology by leveraging three theoretically guaranteed MAB techniques [upper confidence bound (UCB), Thompson sampling (TS), and minimax optimal stochastic strategy (MOSS)] to derive optimal band selection policies. Based on these adopted policies, we propose three algorithms, namely, EABS-UCB, EABS-TS, and EABS-MOSS, to implement the EABS strategy, respectively. Extensive simulations demonstrate our proposed algorithms’ superior performance compared to the traditional link selection schemes regarding energy efficiency, average throughput, and convergence rate. In particular, EABS-MOSS emerges as the best algorithm as it exhibits near-optimal performance due to its flexibility to both stochastic and adversarial environments.

While hybrid-band wireless systems recently gained prominence to achieve high capacity, selecting the best channel in these systems in real-time is still a formidable research challenge that requires further investigations. In this paper, we address this challenge in terms of an optimization problem, which is reformu-lated as a stochastic multi-armed bandit (MAB). Then, we introduce online learning-based solutions to solve the MAB problem for the multi-band/channel selection (MBS). Improved variants of the upper confidence bound (UCB) scheme are investigated and modified to be energy-aware. Hence, we propose Energy-Aware Randomized UCB-MBS (EA-RUCB-MBS) and Energy-Aware Kullback-Leibler UCB-MBS (EA-KLUCB-MBS) methods, which demonstrate near-optimal results. Also, EA-KLUCB-MBS exhibits the fastest convergence, while the convergence of EA-RUCB-MBS is similar to that of the original UCB. Based on extensive simulation results, we evaluate the performance of our proposed algorithms against benchmark MBS schemes including UCB and Thompson sampling (TS).

Relaying is a key technology for millimeter wave communications to extend the range and to route around blockages. In practical implementation of relaying systems, probing is required to identify proper neighbor terminals which will serve as relays. There is however an inherent trade-off between relay probing and required overhead. In this paper, we consider WiGig (IEEE 802.11ad) devices, which are multiband capable with Wi-Fi support, and propose a Wi-Fi assisted relay probing for WiGig device-to-device networks. In the proposed scheme, Wi-Fi received signal strengths are used to pre-select the WiGig relays expected to maximize the spectral efficiency of the overall system. Then, only these pre-selected relays are used in online relay probing step. Simulation analysis demonstrate improvements in throughput and energy consumptions over existing relay probing schemes.

Continuous low-rate monitoring is an important IoT application, which requires high-fidelity in observing signals with low frequency. However, most sensors exhibit noise that is inversely-proportional to spectral frequency (1/f noise). Because both the relevant signal and noise share the same spectral properties, standard linear filtering techniques cannot be used. We are looking into a special application for remote healthcare of the magnetic field sensing of cardiac activity, magnetocardiography (MCG). For such an application, we need to develop a noise separation method, that is also resource-efficient. Previously, we demonstrated AI-based removal of 1/f noise in MCG by a convolutional neural network coupled with gated recurrent units. However, it needs a large amount of data for training, requiring significant training time and computational power. In this work, we employ reservoir computing (RC) for noise-removal, while being conservative in computing resources.

To combat the novel coronavirus (COVID-19) spread, the adoption of technologies including the Internet of Things (IoT) and deep learning is on the rise. However, the seamless integration of IoT devices and deep learning models for radiograph detection to identify the presence of glass opacities and other features in the lung is yet to be envisioned. Moreover, the privacy issue of the collected radiograph data and other health data of the patients has also arisen much concern. To address these challenges, in this paper, we envision a federated learning model for COVID-19 prediction from radiograph images acquired by an X-ray device within a mobile and deployable screening resource booth node (RBN). Our envisioned model permits the privacy-preservation of the acquired radiograph by performing localized learning. We further customize the proposed federated learning model by asynchronously updating the shallow and deep model parameters so that precious communication bandwidth can be spared. Based on a real dataset, the effectiveness of our envisioned approach is demonstrated and compared with baseline methods.

Reporting fine-grained power consumption readings periodically in advanced metering infrastructure (AMI) results in transmitting a massive amount of data by each smart meter (SM). To collect these readings efficiently, change and transmit (CAT) approach can be used. In CAT, the SM sends a consumption reading only when there is enough change in the consumption, which reduces the number of transmitted readings. However, using the CAT approach may trigger attackers to launch a presence-privacy attack (PPA) to infer sensitive information such as the absence of the house occupants by analyzing their SM’s transmission pattern. Therefore, in this paper, we propose a scheme, called “STID”, for collecting the power consumption readings efficiently in AMI networks while preserving the consumers’ privacy by transmitting spoofing transmissions based on an interactive deep-learning defense model. First, we create a dataset that contains the CAT transmission patterns using real power consumption readings and a clustering technique. Next, we train a deep-learning-based attacker model to launch PPA, and the results indicate that the success rate of the attacker is about 90%. Finally, to mitigate the PPA, we train a defense model using deep-learning to transmit spoofing transmissions. The evaluations of our envisioned STID scheme demonstrate a significant reduction in the attacker’s success rate while achieving high efficiency in terms of the number of readings that should be transmitted. Our measurements indicate that our proposed STID can reduce the attacker’s success rate to 6.12% and increase efficiency by about 38% compared to transmitting readings periodically.

In advanced metering infrastructure, smart meters (SMs) send fine-grained power consumption readings periodically to the utility for load monitoring and energy management. Change and transmit (CAT) is an efficient approach to collect these readings, where the readings are not transmitted when there is no enough change in consumption. However, this approach causes a privacy problem, that is, by analyzing the transmission pattern of an SM, sensitive information on the house dwellers can be inferred. For instance, since the transmission pattern is distinguishable when dwellers are on travel, attackers may analyze the pattern to launch a presence-privacy attack (PPA) to infer whether the dwellers are absent from home. In this article, we propose a scheme, called “STDL,” for efficient collection of power consumption readings in advanced metering infrastructure (AMI) networks while preserving the consumers’ privacy by sending spoofing transmissions using a deep-learning approach. We first use a clustering technique and real power consumption readings to create a data set for transmission patterns using the CAT approach. Then, we train a deep-learning-based attacker model, and our evaluations indicate that the attacker’s success rate is about 91%. Finally, we train a deep-learning-based defense model to send spoofing transmissions efficiently to thwart the PPA. Extensive evaluations are conducted, and the results indicate that our scheme can reduce the attacker’s success rate to 3.15%, while still achieving high efficiency in terms of the number of readings that should be transmitted. Our measurements indicate that the proposed scheme can increase efficiency by about 41% compared to continuously transmitting readings.

In advanced metering infrastructure (AMI), smart meters (SMs) are installed at the consumer side to send fine-grained power consumption readings periodically to the system operator (SO) for load monitoring, energy management, and billing. However, fraudulent consumers launch electricity theft cyber attacks by reporting false readings to reduce their bills illegally. These attacks do not only cause financial losses but may also degrade the grid performance because the readings are used for grid management. To identify these attackers, the existing schemes employ machine-learning models using the consumers' fine-grained readings, which violates the consumers' privacy by revealing their lifestyle. In this article, we propose an efficient scheme that enables the SO to detect electricity theft, compute bills, and monitor load while preserving the consumers' privacy. The idea is that SMs encrypt their readings using functional encryption (FE), and the SO uses the ciphertexts to: 1) compute the bills following the dynamic pricing approach; 2) monitor the grid load; and 3) evaluate a machine-learning model to detect fraudulent consumers, without being able to learn the individual readings to preserve consumers' privacy. We adapted an FE scheme so that the encrypted readings are aggregated for billing and load monitoring and only the aggregated value is revealed to the SO. Also, we exploited the inner-product operations on encrypted readings to evaluate a machine-learning model to detect fraudulent consumers. The real data set is used to evaluate our scheme, and our evaluations indicate that our scheme is secure and can detect fraudulent consumers accurately with low communication and computation overhead.

Multihop device-to-device (D2D)-enabled relay networks are envisaged to be utilized by the Internet of Things (IoT) and massive machine-type communication (mMTC) traffic for the purpose of offloading data in beyond fifth-generation (B5G) networks. The emerging challenge of spectrum scarcity and overloading of cellular base stations can be addressed using such relay nodes in terms of spectrum and energy efficiency. In order to improve spectral efficiency, in this article, we intend to employ several frequency bands concurrently in the relay node rather than the traditional concept of specifying one channel on a specific band at a time. A deep learning-based predictive channel selection method is leveraged to unravel the potential challenges associated with the dynamic channel conditions in the multiband relay networks. For predicting the most appropriate channel based on its quality, signal-to-interference-plus-noise-ratio (SINR) is adopted as the metric, which is predicted by the proposed convolutional neural network (CNN) model. The best modulation and coding rates of the predicted band are attained in order to transmit the packets received from the source or previous relay node to the successive relay node/destination. Two proactive channel assignment strategies, referred to as controlled and smart prediction schemes, are employed to exhibit the performance of the shallow and deep-CNN models. The proposed model is evaluated on multiple publicly available data sets from diverse network systems and compared with several machine/deep learning methods. Our proposal leads to encouraging results for proactively predicting the conditions of the channels and choosing the most suitable ones in multiband relay systems.

Purveyors of malicious network attacks continue to increase the complexity and the sophistication of their techniques, and their ability to evade detection continues to improve as well. Hence, intrusion detection systems must also evolve to meet these increasingly challenging threats. Machine learning is often used to support this needed improvement. However, training a good prediction model can require a large set of labeled training data. Such datasets are difficult to obtain because privacy concerns prevent the majority of intrusion detection agencies from sharing their sensitive data. In this paper, we propose the use of mimic learning to enable the transfer of intrusion detection knowledge through a teacher model trained on private data to a student model. This student model provides a mean of publicly sharing knowledge extracted from private data without sharing the data itself. Our results confirm that the proposed scheme can produce a student intrusion detection model that mimics the teacher model without requiring access to the original dataset.

As we are about to embark upon the highly hyped “Society 5.0”, powered by the Internet of Things (IoT), traditional ways to monitor human heart signals for tracking cardio-vascular conditions are challenging, particularly in remote healthcare settings. On the merits of low power consumption, portability, and non-intrusiveness, there are no suitable IoT solutions that can provide information comparable to the conventional Electrocardiography (ECG). In this paper, we propose an IoT device utilizing a spintronic-technology-based ultra-sensitive Magnetic Tunnel Junction (MTJ) sensor that measures the magnetic fields produced by cardio-vascular electromagnetic activity, i.e. Magentocardiography (MCG). We treat the low-frequency noise generated by the sensor, which is also a challenge for most other sensors dealing with low-frequency bio-magnetic signals. Instead of relying on generic signal processing techniques such as moving average, we employ deep-learning training on bio-magnetic signals. Using an existing dataset of ECG records, MCG signals are synthesized. A unique deep learning model, composed of a one-dimensional convolution layer, Gated Recurrent Unit (GRU) layer, and a fully-connected neural layer, is trained using the labeled data moving through a striding window, which is able to smartly capture and eliminate the noise features. Simulation results are reported to evaluate the effectiveness of the proposed method that demonstrates encouraging performance.

Preserving the customers' privacy, while collecting their power consumption for monitoring and billing, is a prime concern in an Advanced Metering Infrastructure (AMI) network of the Smart Grid (SG). In this paper, we address this concern by formally formulating the data aggregation privacy problem, and propose a uniquely crafted Privacy-Preserving Monitoring and Billing scheme using Functional Encryption, referred to as PMBFE. Our proposed PMBFE fulfills four key objectives: (i) data aggregation for billing, (ii) dynamic pricing flexibility, (iii) load monitoring with customers' privacy preservation; and (iv) analysis on how the adopted functional encryption is able to jointly perform data aggregation efficiently and guarantee privacy-preservation. Our envisioned PMBFE approach is evaluated with extensive computer-based simulations. In contrast with the widely employed homomorphic-based encryption in AMI networks, our proposed PMBFE demonstrates significant performance improvement in terms of both communication and computation overheads while guaranteeing user-data privacy. Furthermore, the conducted security analysis exhibits the robustness of our proposal against collusion and eavesdropping attacks.

Traditionally, the Internet of Things (IoT) devices, deployed on the ultra-edge of the network, lack computation, and energy resources. In this paper, we press on the need to go beyond the realms of traditional edge computing (e.g., limited to user-smartphones) and investigate how to incorporate intelligence into the ultra-edge IoT sensors. Among numerous use-cases, we select a mobile Health (mHealth) scenario where we conceptualize a smart IoT sensor to collect and intelligently process single-channel Electrocardiogram (ECG) signals to detect arrhythmia, a heart-condition often associated with morbidity and even mortality. The arrhythmia detection can be regarded as a non-linear Delay Differential Equation (DDE) time-series analysis problem, and the conventional solutions to this problem are not suitable for integration with IoT sensors due to rigorous pre-processing steps. As a solution, a Convolutional Neural Network (CNN)-based, lightweight Arrhythmia classification system is proposed in the paper without the need for noise-filtering and feature extraction steps. Four classes of the heartbeats are considered to comply with the ANSI/AAMI EC57:1998 standard. The proposed system's performances and generalization potential are assessed using three datasets from PhysioNet trained on a deep learning workstation and then transferred to virtualized micro-controllers connected to IoT sensors. The proposed deep learning model exhibits encouraging performance (accuracy 95.27%) in heartbeat classification. Experimental and numerical results demonstrate that the proposed deep learning technique outperforms conventional DDE-based optimization techniques and machine learning techniques such as K-Nearest Neighbor (KNN), and random forest (RF).

Smart grid has emerged as a promising technology for enabling bi-directional communication between the power company and its users to facilitate intelligent, robust, and resilient next generation power grid systems. Through this technology, both the power company and its subscribers can be equally benefited, not only from economic point of view, but also in terms of environment-friendly quality of power usage. One important challenge for the smart grid designers is the demand side management, which can lead to avoiding the peak hours and reducing the cost for the consumers. In this paper, we address the power balancing challenge for the smart grid and discuss different solutions including game theoretic methods and demand control policies. Also, we present our novel demand control policy for achieving an effective management of the power consumption. Computer simulations demonstrate the effectiveness of the proposed policy compared to existing ones.

Due to their ability to overcome many shortcomings associated with the contemporary client-server paradigm, Peer-to-Peer (P2P) networks have attracted phenomenal interests from researchers in both academia and industry. Interactive and multimedia streaming applications using P2P networks are, however, often prone to long startup delays, which disrupt the smooth playback and undermine users' perceived quality of service. In addition, P2P networks must be able to support a potential number of users while ensuring that the resources are efficiently utilized. In this paper, by addressing these shortcomings in the traditional P2P framework, we envision a novel scheme to effectively provide a Video-on-Demand (VoD) using P2P-based mesh overlay networks. The proposed scheme covers two main phases, namely requesting and scheduling modes. The former aims at dynamically selecting the required contents from the available peers. On the other hand, in the scheduling mode, the incoming requests are scheduled in a priority-based manner for minimizing the startup latency and sustaining the playback rate to an acceptable level. Computer simulations have been conducted to verify the effectiveness of the proposed scheme. The obtained results demonstrate the scalability of our envisioned scheme in addition to its capability to reduce the startup delay and provide a sustainable playback rate.