Practical Improvements of Profiled Side-Channel Attacks on a Hardware Crypto-Accelerator

Abstract

This article investigates the relevance of the theoretical framework on profiled side-channel attacks presented by F.-X. Standaert et al. at Eurocrypt 2009. The analyses consist in a case-study based on side-channel measurements acquired experimentally from a hardwired cryptographic accelerator. Therefore, with respect to previous formal analyses carried out on software measurements or on simulated data, the investigations we describe are more complex, due to the underlying chip’s architecture and to the large amount of algorithmic noise. In this difficult context, we show however that with an engineer’s mindset, two techniques can greatly improve both the off-line profiling and the on-line attack. First, we explore the appropriateness of different choices for the sensitive variables. We show that a skilled attacker aware of the register transfers occurring during the cryptographic operations can select the most adequate distinguisher, thus increasing its success rate. Second, we introduce a method based on the thresholding of leakage data to accelerate the profiling or the matching stages. Indeed, leveraging on an engineer’s common sense, it is possible to visually foresee the shape of some eigenvectors thereby anticipating their estimation towards their asymptotic value by authoritatively zeroing weak components containing mainly non-informational noise. This method empowers an attacker, in that it saves traces when converging towards correct values of the secret. Concretely, we demonstrate a 5 times speed-up in the on-line phase of the attack.