With the deepening of research into chaotic systems, chaotic block ciphers have found widespread application in multiple fields such as the Internet of Things. Currently, the security of chaotic block ciphers has been validated through traditional statistical analysis. However, during their hardware implementation, chaotic block ciphers may exhibit physical information leakage, making them susceptible to side channel analysis that severely threatens the security of the ciphers. To further enhance the security analysis of chaotic block ciphers, this paper investigates chaotic block ciphers based on the Feistel architecture from an attack perspective. Compared with traditional attack methods — such as template attacks that suffer from singular matrices and low attack efficiency — this paper proposes a neural network-based side channel analysis method. This approach effectively avoids the occurrence of singular matrices during the process of template attacks and improves attack efficiency. Meanwhile, to further enhance attack performance, this paper proposes side channel analysis methods based on neural networks and Support Vector Machine (SVM) that utilize heuristic optimization algorithms. This approach mainly consists of three stages: model parameter optimization, training, and attack execution. Experimental analysis reveals that the optimized neural network can recover the algorithmic key using only 40 energy traces. Meanwhile, the correlation coefficient of the optimized SVM for guessing the correct key reaches approximately 0.95. It can identify the correct key with fewer than 20 energy traces, achieving a success rate of 100% and a guessing entropy approaching 0. In summary, attack methods based on optimization algorithms can effectively improve the success rate of key guessing, avoid singular matrices, and thus enable effective analysis of the hardware security of chaotic block cipher systems.
Xi et al. (Fri,) studied this question.