Revisiting System Noise in Side-Channel Attacks: Mutual Assistant SCA vs. Genetic Algorithm
Rei Kudo, Takeshi Sugawara, Kazuo Sakiyama, Yuko Hara–Azumi, Li Yang · 2021
For side-channel attacks (SCA) against parallel S-boxes such as AES, side-channel leakage that is not related to the targeted byte becomes a system noise for the key recovery of the target byte. With a recovered key byte for an S-box calculation, one can mitigate the corresponding system noise to assist the recovery of other key bytes. Following this idea, this paper proposes a mutual assistant side-channel attack (MA-SCA) that improves SCA efficiency using partially recovered keys. MA-SCA is inspired by the simple genetic algorithm CPA (SGA-CPA) and multiple sieve CPA (MS-CPA) attack that combines genetic algorithm with correlation power analysis. In genetic algorithm-based CPA, the partial key pieces are randomly combined and sieved repeatedly to solve the optimization problem of the correlation coefficient. In MA-SCA, the partial key pieces are combined in an organized manner to achieve a compact balance between the computation and the data complexities. In the evaluation, compared with MS-CPA, mutual MA-CPA archived a similar success rate using only 1–5% of the calculation. This work also discusses the boundaries of SCA improvement by reducing system noise.