This paper presents RIS-FL-HO-V2X, a privacy-preserving framework that tightly integrates federated reinforcement learning (FRL) and reconfigurable intelligent surfaces (RISs) to jointly optimize RIS-aware handover decisions in dynamic urban vehicle-to-everything (V2X) communications. The proposed solution addresses urban blockage, severe Doppler effects, imperfect channel state information (CSI), and abrupt spatial-temporal fluctuations in base station (BS)/vehicle density by exploiting RIS-enabled line-of-sight (LoS) enhancement and a shared federated handover policy trained from partial local observations while keeping raw data on vehicles. In the proposed architecture, vehicle users execute RIS-aware handover decisions based on local observations, while their model updates are securely aggregated at the macro base station, which also centrally optimizes RIS phase-shift configurations. A carefully designed RIS-aware multi-objective reward balances throughput, delay, reliability, handover and vertical handover costs, packet loss, outage, and RIS reconfiguration overhead. Simulation results show that RIS-FL-HO-V2X outperforms centralized RL, fully distributed RL, and non-learning baselines by substantially reducing vertical handover frequency, outage probability, packet loss ratio, handover delay, and RIS variation rate, while significantly improving experienced signal-to-interference-plus-noise ratio (SINR), LoS probability, and overall V2X throughput.
Traditional minimum-cost spanning tree methods for the general edge coverage problem tend to overprotect redundant edges, raising costs and complexity when applied to large networks such as power grids, fiber-optic backbones, and IoT infrastructures. We introduce two concepts: the essential graph that encodes necessary connectivity and the edge perimeter that separates protected from unprotected edges. Our algorithm uses these concepts to eliminate redundant edges while preserving connectivity, dramatically improving efficiency. Experiments demonstrate a million-fold speedup over exact algorithms. For tiny instances, the approach achieves perfect precision with an average extra cost below 1
Internet of Medical Things (IoMT) has completely changed healthcare system through real-time diagnosis and remote monitoring of patients. However, medical data is sensitive to sophisticated cyber threats; thus, protecting patient data is still a major concern. Traditional Intrusion Detection Systems (IDSs) are often unable to detect unrecognized attacks leading to compromised data security and high false alarm rates, and traditional machine learning (ML) models struggle with complexity of patterns in IoMT data. Hence, this study presents a novel privacy preserving Cross Attention based Adaptive Quantum Convolutional Neural Network (CA-AQCNN) approach for efficient Cyberattack detection in IoMT applications. Gathered IoMT data stored in distributed cloud nodes using a novel privacy-aware framework that offers fusion of heterogeneous data by employing deep learning (DL), while preserving privacy of data. A novel Deep Residual Capsule Autoencoder (DResCapsAE) is introduced to fuse IoMT heterogeneous data and input to quantum neural network, which is trained to detect cyberattacks in IoMT systems. Experimental results demonstrate that the proposed system attained 99.89
Wireless Sensor Networks (WSNs) provide reliable connectivity and data-gathering capabilities, but accurate node localization remains a persistent challenge, especially in 3D environments where altitude variations complicate distance estimation. Most existing 3D localization methods rely on a minimum of four anchor nodes (ANs), which significantly increases deployment costs. To reduce this cost, a single-anchor 3D localization approach is proposed that uses a virtual-anchor projection strategy combined with a Lévy Flight-Enhanced Adaptive Butterfly Optimization Algorithm (LF-EABOA) to improve accuracy and efficiency. As soon as a mobile target node falls within the communications range of a single anchor node, six virtual anchor nodes at the same distance are created around the anchor node. After that, an adaptive selection mechanism will select the best virtual anchors that will provide robust localization without the need for deployment of physical anchors. This method explicitly models noise and interference using a log-distance path-loss model, including shadowing effects and a degree of irregularity (DOI) and non-line-of-sight (NLOS) penalties. In order to verify the proposed method, extensive simulation experiments were conducted using an average of 80 nodes to represent medium-scale configurations, which demonstrated that the proposed LF-EABOA method produced a mean localization error of 0.1895 m, which was significantly less than existing traditional methods and was also associated with substantially lower costs. Statistical analysis of confidence intervals and mobility sensitivity demonstrate that the proposed method is robust.
The increasing adoption of smart home technologies has intensified the need for intelligent and privacy-preserving Human Activity Recognition (HAR) systems. Centralized learning approaches expose sensitive behavioral data to privacy risks, while existing federated learning methods remain vulnerable to inference attacks and face challenges arising from heterogeneous and resource-constrained home environments. To address these limitations, we propose FedPrivHAR-LLM, a unified framework that integrates federated learning, differential privacy, homomorphic encryption, and Large Language Models (LLMs). Federated learning enables distributed model training without raw data sharing, differential privacy provides formal statistical guarantees against re-identification, and homomorphic encryption secures model aggregation in the encrypted domain. Beyond secure training, LLM-based semantic reasoning is incorporated to enhance context-aware interpretation and disambiguation of human activities. Experimental evaluations on three benchmark HAR datasets adapted to smart-home settings (CASAS, Opportunity, and PAMAP2) show that FedPrivHAR-LLM achieves up to 90.2
