By introducing collision information, the existing side-channel Correlation-Enhanced Collision Attacks (CECAs) performed collision-chain detection, quickly filtered out candidates unsatisfying collision conditions and extracted a part of optimal candidates for further process, thereby rapidly and significantly reducing the key candidate space and the difficulty of key recovery. However, they are still limited by disadvantages such as serial implementation, complex parameter settings and lack of intelligence, resulting in a low success rate of key recovery. To address these issues, we first present a Collision Detection framework with Genetic Algorithm (CDGA), which exploits Genetic Algorithm to detect the collision chains and has a strong capability of global searching. Secondly, we theoretically analyze the performance of CECA, and bound the searching depth of its output candidate vectors with a confidence level using a data-driven hypothesis test that provides confidence bounds for Gaussian leakages and an approximation based on Central Limit Theory (CLT)for non-Gaussian cases, which facilitates effective and stable population initialization. Thirdly, benefiting from our hypothesis-test-guided design, we propose a goal-directed mutation that prioritizes promising collision candidates, thus improving efficiency and adaptability of the CDGA. Finally, to optimize the evolution of CDGA, we introduce a roulette selection strategy to employ a probability assignment based on individual fitness values to guarantee the preferential selection of superior genes. Comprehensive experiments on DPA Contest v4.1 (AES-256 with Rotated S-boxes Masking) and an AT89S52 AES-128 platform demonstrate that CDGA achieves faster convergence and higher key-recovery success rates compared with TOC/FTC/FCC and Wiemers’ cumulative-correlation selection.

A cryptographic implementation produces very similar power leakages when fed with the same input. Side-channel collision attacks exploit these similarities to establish the relationship between sub-keys and improve the efficiency of key recovery. Benefiting from independence of leakage model, they play an important role in non-profiled setting. However, performance of existing approaches against single collision value is still sub-optimal and optimization is promising. Motivated by this, we first theoretically analyze the mathematical dependency between the number of collisions and the number of encryptions, and propose an efficient side-channel attack named Collision-Paired Correlation Attack (CPCA) to guarantee that the side with fewer samples in a collision is completely paired in low noise scenario. This allows overcoming the inefficient utilization of information in existing works. Moreover, to further employ underlying informativeness, we maximize collision pairs as many as possible. This optimization significantly improves performance of CPCA and thereby extends it to large noise scenarios. Finally, to achieve moderate computational complexity, two equivalent variants of CPCA are investigated to address the potential problem of limited computing resources. Our further theoretical study illustrates that CPCA provides the upper security bound of Correlation-Enhanced Collision Attack (CECA), and experimental results fully verify its superiority.

Evaluating cryptographic implementations with respect to side-channel analysis (SCA) has been mandated at high security levels. Typically, the evaluation involves four stages: detection, modeling, certification and recovery. In pursuit of a specific goal at each stage, inherently different techniques were previously considered necessary. However, since the recent Eurocrypt 2022 and Eurocrypt 2024, linear regression analysis (LRA) has become the unique technique well-applied throughout all the stages. In this paper, we concentrate on this “silver bullet” technique within the field of SCA. In the first part of this paper, we answer three fundamental questions organized progressively. The first one relates to “why use LRA?”. Our discussion of the nominal and binary nature elucidates its critical role in underpinning the state-of-the-art techniques. Having understood the merits, a natural follow-up is “how to use it (correctly and effectively)?”. A theoretical analysis of the design matrix is provided, regarding the sample distribution of plaintext and the chosen degree of polynomial. We summarize the conditions for eliminating multicollinearity, a problem that can be harmful to all LRA-based techniques. The last question “who should use LRA?” reveals an intriguing evaluator-advantageous property: LRA can only unleash its full potential when the key is known. In the second part of this paper, we clarify the connections between LRA and traditional SCA techniques. Our proofs provide new insights into the prior investigation of SCA reduction, fostering a comprehensive understanding of this linear family. The conclusions suggest that the core working mechanisms of the state-of-the-art techniques can be traced back to those of earlier differential side-channel analyses. Experimental results are in line with the theory, confirming its correctness in practice.

Success Rate (SR) is one of the most popular side-channel security metrics measuring the efficiency of key recovery. Theoretical expression of success rate reveals the functional dependency between relevant parameters such as number of measurements and Signal-to-Noise Ratio (SNR), helping researchers understand the resistance of a given implementation rapidly. However so far, existing works have exposed fundamental problems: (1) Evaluation is confined to a very limited range of distinguishers and specialized methods; (2) Evaluation assumes a perfect leakage model that is detached from reality. It is widely observed that an inaccurate leakage model will lead to a degraded or even distorted success rate. In this paper, we tackle above problems by introducing a novel framework which is able to evaluate seven side-channel distinguishers with a unified expression. Among them, we explore four new distinguishers that have not been investigated in the existing literature. Within the framework, DPA distinguishers are intuitively understood as linear maximum likelihood attack testing closeness between vectors with some easy-to-comprehend geometric metrics. Our evaluation is able to deal with profiled models of any quality and is agnostic to model profiling techniques. It uniquely enables the evaluation of success rates under inaccurate leakage models, whilst providing an (indirect) answer to the open question “how much information is lost due to the model biases” through quantifying the degradation of success rates. Finally, we formulate a set of criterion values for quantitative analyses of the model biases. It provides theoretical evidences for a more thorough explanation for the various behaviors of DPA attacks. Experimental results are inline with the theory, confirming its practical applicability.