Filtering-Based Correlation Power Analysis (CPA) with Signal Envelopes Against Shuffling Methods

Abstract

The Correlation Power Analysis (CPA) is one of the powerful Side-Channel Analysis (SCA) methods to reveal the secret key using linear relationship between intermediate values and power consumption. To defense the analysis, many crypto-systems often embed the shuffling implementation which shuffles the order of operations to break the relationship between power consumption and processed information. Although the shuffling method increases the required number of power traces for deploying the CPA, it is still vulnerable if an attacker can classify or group the power traces by operations. In this work, we propose a new CPA technique by efficiently clustering the power traces using signal envelopes. We demonstrate theoretically reduced time complexity and tested our approach with the eight-shuffling AES implementations.