Key points are not available for this paper at this time.
Smart consumer devices are increasingly integrated with Internet of Things (IoT) and Augmented/Virtual Reality (AR/VR), including holographic interfaces, making them targets for sophisticated cyber-attacks. Despite incorporating advanced biometric authentication systems, the rapid growth in personal data transmission over wireless networks invites various zero-day and unknown threats. To address this issue, cognitive digital forensic systems are crucial but face challenges in real-world deployment due to high communication latency, network overhead, and security risks during multi-source data collection. This paper proposes a novel Federated U-Siamese Swin Transformer Networks framework to enhance device-level security. On the client side, a dual U-Siamese Swin V2 model with zero stopping gradients is utilized for accurate and lightweight biometric verification. On the server side, a dynamic federated aggregation strategy is employed to ensure robust and privacy-preserving learning. This framework enables decentralized yet collaborative detection of zero-day attacks while minimizing communication cost. Experimental validation using the LiveDet dataset demonstrates significant improvements in performance, achieving an accuracy of 0.994, precision of 0.992, recall of 0.990, specificity of 0.990, and F1-score of 0.992. When evaluated against state-of-the-art methods, the proposed framework demonstrates significant improvements in detection speed and communication efficiency, aligning with the demands of real-time systems.
Alqhatani et al. (Mon,) studied this question.