A Survey of Methods and Architectures for Secure AI for 6G Mobile Devices: Deep Kronecker Neural Network Optimized with Hybrid Cat Hunting Optimization to Combat Side-Channel Attacks

Abstract

The emergence of sixth-generation (6G) mobile networks introduces unprecedented capabilities such as ultra-low latency, massive connectivity, and intelligent automation. However, these advancements also expose mobile devices to sophisticated security threats, particularly side-channel attacks that exploit physical leakages like power consumption, timing, and electromagnetic emissions. This paper presents a comprehensive survey of methods and architectures for secure artificial intelligence (AI) in 6G mobile environments, with a focus on deep learning-based defenses. Specifically, the study explores the integration of Deep Kronecker Neural Networks (DKNN) with Hybrid Cat Hunting Optimization (HCHO) to enhance model robustness and computational efficiency. Recent advancements between 2020 and 2023 are analyzed, highlighting the evolution of convolutional neural networks (CNNs), recurrent neural networks (RNNs), and hybrid architectures in detecting and mitigating side-channel vulnerabilities. Comparative analysis reveals that optimized deep learning models can achieve detection accuracies exceeding 95% while maintaining low computational overhead. The paper also identifies challenges such as model generalization, adversarial resilience, and real-time deployment constraints. The proposed framework demonstrates the potential of combining advanced neural architectures with metaheuristic optimization to ensure secure and efficient AI-driven 6G mobile systems.