More Practical and Robust: Enhancing Simple Power Analysis on Cryptosystems with Double Clustering

The widespread use of public key cryptographic algorithms in embedded devices has made them a primary target for side-channel analysis (SCA). Clustering-based simple power analysis (SPA) poses a significant threat to public key implementations by inferring secret keys through the identification of distinguishable patterns in side-channel information. However, the traditional clustering-based SPA methods are highly sensitive - even to non-key-dependent patterns - thereby limiting their robustness and practical applicability. To address these limitations, this article proposes a double-clustering method that enhances the flexibility, accuracy, and robustness of clustering-based SPA. By progressively adjusting the number of clusters, the method adaptively identifies optimal clustering configurations, mitigating the need for fixed assumptions and improving resistance to noise and other interfering factors. Experiments covering multiple cryptographic algorithms, hardware platforms, and countermeasure settings demonstrate that the proposed method consistently outperforms traditional clustering-based SPA methods.