Paradigm Shift of Machine Learning to Deep Learning in Side Channel Attacks - A Survey

Abstract

A side-channel attack is a sort of computer security attack depending on data obtain from program implementation rather than flaws in the program itself. Additional data sources that might be exploited include timing data, power consumption, electromagnetic leakage, and even sound. Since the 1990's, when side channel attacks were first introduced, a lot of effort is done in improving their effectiveness and efficiency. Many machine learning methods were designed for side-channel attacks. The researchers have recently noticed a growing trend in the SCA community to use deep learning techniques, which has resulted in more precise side-channel studies, even when countermeasures are in effect. Machine learning algorithms, on the other hand, have the drawback of requiring human engineering to function and some performance fluctuations in some cases. Recent research has focused on using deep learning techniques to extract characteristics from data automatically. Concepts of side channel attacks, machine learning, deep learning, and current breakthroughs in deep learning-based side-channel attacks are described in this paper. This discussion provides an overview of contemporary machine and deep learning research in the context of SCA. Machine Learning is not completely negated, or Deep Learning is not completely supported but their competitiveness and trend is discussed as a baseline to new challenges in context of side channel attack. Thus, a paradigm shift of ML to DL in SCAs is the focus of this paper.