Side-channel attacks and machine learning approach

Abstract

Most modern devices and cryptoalgorithms are vulnerable to a new class of attack called side-channel attack. It analyses physical parameters of the system in order to get secret key. Most spread techniques are simple and differential power attacks with combination of statistical tools. Few studies cover using machine learning methods for pre-processing and key classification tasks. In this paper, we investigate applicability of machine learning methods and their characteristic. Following theoretical results, we examine power traces of AES encryption with Support Vector Machines algorithm and decision trees and provide roadmap for further research.

Keywords

• Side-channel attacks
• AES
• Machine learning