A study on clustering-based power trace classification for side channel attacks

Template attack (TA) is an effective method to perform side channel attacks, which assumes that the attacker has the full control of a cryptographic device before attacking a similar one. In this paper, a TA method using semi-supervised learning strategy is used to relax this assumption. The effectiveness of our proposed method is confirmed by software simulations, in which the data are collected from the experiment that conducted on an 8 bit microcontroller. From the comparison that made between two supervised machine learning methods and the proposed one, it shows that the method in this paper has a higher success rate of key recovery.