Parala Maharaja Engineering College

Multiplex graphs capture diverse relations among shared nodes. Most predictors either collapse layers or treat them independently. This loses crucial inter-layer dependencies and struggles with scalability. To overcome this, we frame multiplex link prediction as multi-view edge classification. For each node pair, we construct a sequence of per-layer edge views and apply cross-layer self-attention to fuse evidence for the target layer. We present two models as instances of this framework: Trans-SLE, a lightweight transformer over static embeddings, and Trans-GAT, which combines layer-specific GAT encoders with transformer fusion. To ensure scalability and fairness, we introduce a Union–Set candidate pool and two leakage-free protocols: cross-layer and inductive subgraph generalization. Experiments on six public multiplex datasets show consistent macro-F_1 gains over strong baselines (MELL, HOPLP-MUL, RMNE). Our approach is simple, scalable, and compatible with both precomputed embeddings and GNN encoders.

Federated learning (FL) approach plays a vital role in reducing privacy risks by keeping raw data localized at clients however, various attacks including model inversion or inference demonstrate that transmitting local model updates $\left(f w_{i}\right)$ can still leak sensitive information. The objective of this work is to address the critical scalability barrier impeding the widespread deployment of privacy-preserving FL. Secure Aggregation (SA) protocols address this vulnerability by ensuring that the aggregation server only observes the sum of client updates. Current SA scheme, such as SECAGG, have prohibitive computational complexity at the central server, specifically $O\left(L N^{2}\right)$ which limits their applicability in large-scale FL deployments. In this paper, we implemented and evaluated the FastSecAgg secure aggregation protocol for medical FL and targeting communication and computational efficiency. FastSecAgg deploys an FFTbased multi-secret sharing algorithm, and FastShare to reduce the computational complexity of $O(L \log N)$ under a honest-but-curious server model. The experimental results showed that the performance under non-IID clinical data distributions and partial update availability, integrating FastSecAgg within Flower FL framework enables privacy-preserving aggregation maintaining model accuracy comparable to non-secure federated averaging. This work provides practical insights into the performance-privacy trade-offs of secure aggregation in medical federated learning.

Abstract A new control approach for shunt active power filters (SAPFs) is proposed in this paper, specifically designed to improve power quality and operational stability in bipolar high voltage direct current (HVDC) transmission systems. It offers dual benefits: enhanced harmonic suppression and a more optimized AC yard structure due to reduced requirement for supplementary reactive power compensators, lowering the HVDC project cost. The novel filter system features a three-level voltage source converter and an appropriate control scheme that integrates synchronous reference frame (SRF) control strategy and self-tuning using high selectivity filter (HSF) that is competent to simultaneously accomplish harmonic suppression and dynamic reactive power support, thereby enhancing the accuracy and dynamic response time, without requirement of additional phase-locked loop. The MATLAB simulation results were obtained for system performance under uncompensated condition and subsequently compared with compensated system employing passive power filter, conventional SRF based SAPF and proposed HSF-SRF based SAPF. The proposed method achieved remarkable reduction in harmonic distortion, with bus voltage total harmonic distortions (THDs) decreasing from approximately 71%–73% in the uncompensated condition to nearly 0.8%, while bus current THDs are reduced from around 20%–22% to nearly 2% using the proposed scheme, under both steady state and sudden load change scenarios, as simulated and demonstrated. This ascertains the advantages of proposed HSF-SRF strategy for SAPF system in bipolar HVDC-links in terms of reactive power compensation and harmonic mitigation, satisfying IEEE-519 standard recommendations on harmonic limits.

The execution of image processing tasks on edge and IoT devices is always coupled with the need to develop encryption algorithms with a guarantee of security and low computation complexity. Most classical encryption algorithms employ a constant secret key and software-based implementation, making them unsuitable for resource-constrained devices and platforms. The study proposes the idea of implementing edgedriven key generation for hardware-assisted lightweight image encryption with Sobel edge detection on the PYNQ-Z2 System on a Chip. The Sobel filter algorithm is implemented in the FPGA hardware for the detection of structural edge characteristics in a real-time process, and the statistical analysis of the resulting edge map is implemented in software to develop dynamic imagerelated encryption keys, with the keys further employed in a lightweight symmetric encryption technique implemented with no reliance on any cryptography software library. The experimental outcome has efficiently shown the best performance on the PYNQ-Z2 platform, confirming the usability of the proposed technique in secure image processing of IoT edge devices.