Bypass Analysis Based on Convolutional Neural Networks

Abstract

Side-channel analysis, as a key branch of cryptanalysis, poses a serious threat to the security of embedded smart devices. With the development of artificial intelligence technology, side-channel cryptanalysis methods based on convolutional neural networks (CNNSCA) have become a research hotspot. This paper focuses on the modeling and analysis of CNNSCA, exploring its application in side-channel cryptographic attacks. In traditional side-channel analysis techniques, non-modeling methods such as Differential Power Analysis (DPA) and Correlation Power Analysis (CPA) require domain knowledge to construct guess models; modeling methods like Template Attacks (TA) involve template construction and matching phases. CNNSCA leverages the advantages of deep learning to automatically extract data features. Experimental results demonstrate that CNNSCA outperforms traditional TA when attacking ideal low-noise side-channel signals, and shows greater advantage in attacking distorted side-channel signals due to noise and countermeasures, providing new avenues for improving the efficiency of side-channel cryptographic attacks and countering protective strategies, which is of significant importance for ensuring the security of cryptographic systems.