The IoT Smart grid systems require privacy-sensitive, scalable, and secure communication, but the traditional cryptographic methods are susceptible to quantum attacks, central failure, and high computational costs. In this paper, the author presents a multi-layered structure that uses various yet complementary approaches in order to solve these issues. Quantum Key Distribution (QKD) is applied to create quantum-resilient key management to provide the secure transmission against future quantum attacks. Homomorphic Encryption (HE) allows the computations to be performed on encrypted data to preserve privacy and minimize the risks of exposure at some point during the processing. Federated Learning (FL) is used to assist the decentralized intelligence, where anomaly detection models are trained together without the need of transferring sensitive raw data. The combination of the two techniques produces a hybrid architecture that is both highly secure and scalable. Experimental validation on real-world smart grid data indicates the 42% communication overhead reduction versus traditional encryption, a 35% improvement in anomaly detection with FL and strong resistance to quantum attack and side-channel attack simulation. The findings support the fact the suggested solution provides end-to-end confidentiality, integrity, and performance efficiency, which will form a basis of quantum-resilience and privacy-preserving communication in future smart grid IoT network.

Early and reliable diagnosis of skin cancer from dermoscopic images remains challenging due to class imbalance, subtle inter-class variations, lesion boundary ambiguity, and illumination inconsistency, which can degrade the robustness of conventional convolutional neural networks (CNNs). To address these limitations, this study proposes an automated smart healthcare framework for dermoscopic skin cancer diagnosis using an Enhanced Vision Transformer (E-ViT) that improves global-context modeling through self-attention while strengthening fine-grained lesion representation learning. Unlike standard ViT configurations, the proposed architecture integrates multi-scale patch embedding and attention refinement to better capture border irregularities and color–texture heterogeneity that are critical for melanoma discrimination. Furthermore, to eliminate manual hyperparameter tuning and ensure stable convergence across imbalanced datasets, we introduce a Dimension Learning–Based Hunting (DLH) Improved Grey Wolf Optimization (IGWO) strategy as the sole hyperparameter optimization engine. DLH-IGWO adaptively balances exploration and exploitation by updating search directions across dimensions, enabling efficient discovery of optimal training settings (e.g., learning rate, batch size, patch size, transformer depth, attention dropout, and weight decay) under constrained computational budgets. Experimental results on the HAM10000 and ISIC-2019 datasets show that the proposed DLH-IGWO–E-ViT consistently outperforms strong CNN and ViT baselines, achieving 98.37% accuracy, 97.78% F1-score, and 99.12% AUC on HAM10000 and 97.21% accuracy with 95.93% macro-F1 on ISIC-2019. The model significantly improves melanoma detection by reducing the false-negative rate to 3.16%, with all gains being statistically significant (p < 0.001) and computationally feasible for clinical deployment.

High-Performance Computing (HPC) systems have become integral to scientific discovery and Artificial Intelligence (AI), yet their complex, distributed architectures expose them to sophisticated cyber threats capable of disrupting large-scale computational workflows. Traditional intrusion detection systems (IDS) lack the adaptability and contextual understanding required to secure such dynamic environments. This study presents an advanced cybersecurity framework that leverages a Graph-Temporal Transformer Network (GTTN) combined with a Hybrid Adaptive Meta-Optimization (HAMO) algorithm to enhance anomaly detection and classification in HPC infrastructures. The proposed framework first constructs graph representations of inter-node communications and process-level dependencies, capturing both structural and temporal relationships within HPC traffic. A self-supervised pretraining phase is employed to learn normal behavioral patterns without labeled data, improving the model’s resilience to zero-day attacks. The fine-tuned GTTN then integrates spatial attention and temporal encoding to detect deviations indicative of intrusion activities. To further optimize model performance, the HAMO algorithm, which fuses the exploration capability of the Spotted Hyena Optimizer (SHO) with the exploitation strength of the Whale Optimization Algorithm (WOA), is applied to adaptively tune the network’s hyperparameters. Experimental evaluations conducted on HPC-tailored versions of the CIC-IDS 2018 and LANL Cyber dataset reveal that the proposed system achieves detection accuracies of 0.995 and 0.982, outperforming state-of-the-art deep learning (DL) and evolutionary baselines. The framework also demonstrates reduced false alarm rates and superior scalability under high-throughput workloads. These results indicate that the integration of graph-temporal learning and hybrid meta-optimization provides a powerful and adaptive defense strategy for next-generation HPC cybersecurity, offering practical potential for deployment in exascale computing environments.