Dual-Path Hybrid Residual Network for Profiled Side-Channel Analysis

• Published in: IEEE transactions on circuits and systems. II, Express briefs Vol. 71; no. 8; pp. 3985 - 3989
• Main Authors: Sun, Shaofei, Yu, Hongwei, Wang, An, Wei, Congming, Ding, Yaoling, Zhu, Liehuang, Chen, Jiazhe
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.08.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Convolution; cryptographic chips; Cryptography; Deep learning; DH-HEMTs; Dual-path; Feature extraction; Neural networks; Performance evaluation; Residual neural networks; Security; side-channel analysis; State-of-the-art reviews; Tuning
• ISSN: 1549-7747; 1558-3791
• DOI: 10.1109/TCSII.2024.3371110

Side-channel analysis poses a significant security threat to cryptographic chips in embedded devices. The use of deep learning in side-channel analysis makes it easier to compromise the security of cryptographic chips. Although these chips equipped with countermeasures can increase the complexity of side-channel analysis, it is essential to continue exploring and developing more advanced analysis methods for better security. In this brief, we propose a simple residual network called ResNet-S, which has shown strong performance. Based on this foundation, we have developed the dual-path hybrid residual network. The dual-path hybrid convolution technique is used for feature fusion. It utilizes a multi-scale convolution strategy to effectively reduces the required number of traces for key recovery. We have evaluated the performance of our proposed neural networks on different datasets, and the experimental results show that our proposed networks outperform the state-of-the-art neural networks that have been published.