As 6G mobile networks evolve, they become increasingly vulnerable to side-channel attack (SCA), which exploit data leaks that occur during communication and processing. The existing detection techniques of side-channel assaults in 6G mobile devices is complicated by issues like noise interference, high-dimensional data complexity, processing time delays and model complexity errors. In this paper, the novel approach proposed for SCA detection in 6G mobile devices using Attributed Multi-order Graph Convolutional Network (AMGCN) optimized with Snow Ablation and Wolf-Bird Optimization. At first, the input data are subject to preprocessing which is gathered from CHES CTF database and eliminate the unnecessary noise using Iterated Rational Quadratic Kernel-high-order Unscented Kalman Filter (IRQ-UKF). After that, the preprocessed data are given to AMGCN classifier for detecting SCA but effectiveness was affected due to high latency. Therefore, the weight is initialized by Wolf-Bird Optimization (WBO) and has the problem of premature convergence due to random selection in search region. Finally, Snow Ablation Optimization (SAO) is included for finding global dependencies and it enhances the AMGCN for accurate detection of SCA. The effectiveness of the proposed approach achieves higher accuracy, precision and RoC of 16.28%, 30.78%, and 25.29% as comparing with existing approaches.