Logistic regression-based side-channel analysis attacks

Abstract

Advanced Encryption Standard (AES) is a modern concept in cryptography. Most of modern encryption schemes and devises are built based on this standard. Those encryption systems have really high resistance to most of modern attacking methods. However, this passage will introduce the most powerful way of attacking: Side-Channel Analysis (SCA). By performing such attacking by artificial intelligence model, it shows that they can break the encryption system in a very efficient and effective way. By using ASCAD database, this passage analysis some properties when using logistic regression to perform Side-Channel-Attack on the traces of the encryption system. Since in the original article, the authors only analysis the performance of multilayer perceptron (MLP) and convolution neural network (CNN), this passage aims to apply similar methodology to logistic regression and analyse its performance in different circumstances. Moreover, some interesting properties about logistic regression was found, and it can sometimes perform better than systems in the original passage in certain situation.