Paper 2023/021

DLFA: Deep Learning based Fault Analysis against Block Ciphers

Abstract

Previous studies on fault analysis have demonstrated promising potential in compromising cryptographic security. However, these fault analysis methods are limited in practical impact due to methodological constraints and the substantial requirement of faulty information such as correct and faulty ciphertexts. Additionally, while deep learning techniques have been widely applied to side-channel analysis (SCA) in recent years and have shown superior performance compared with traditional methods, there has been minimal research focusing on the application of deep learning techniques in fault analysis to date. This paper proposes an innovative approach named deep learning-based fault analysis (DLFA) by incorporating deep learning techniques into fault analysis, which enhances the efficiency and versatility of the analysis. DLFA is equipped with a novel feature selection method to extract valuable information for neural networks from faulty ciphertexts. Besides, optimized hyper-parameters for MLP and CNN are presented as the benchmarks. Experimental results on advanced encryption standard (AES) reveal that DLFA achieves outstanding performance. Notably, DLFA requires only 683 minimum and 1488 average ciphertexts with an average analysis time of 0.12s, surpassing previous works in terms of efficiency.