Investigação do impacto de fenômenos de variabilidade e do efeito BTI na robustez de contramedidas a ataques DPA/DEMA

Abstract

Side-channel attacks consist of a class of techniques that exploit vulnerabilities emerging from the physical implementation of a cryptographic system to extract confidential data. Despite the supposed security provided by cryptographic protocols, side-channel attacks benefit from the inherent correlation between data processed by a system and the physical properties propagated by it. Aiming to neutralize the action of these attacks, countermeasures are designed intending to thwart the correlation. Although they provide a significant reduction in data dependency, this correlation cannot be completely removed. Additionally, the downscaling of CMOS technology results in the increase and relevance of factors of variability and aging on parameters that interfere with integrated circuits. Thus, it intensifies the need for circuits capable of ensuring the confidentiality of the information and offering higher reliability at the system level. Under this scenario, this work proposes to investigate the impact of variability phenomena and the BTI effect in the security provided by circuits designed to prevent side-channel attacks. The obtained results demonstrated that process variability severely affects the protection provided by countermeasures, substantially reducing their effectiveness, and should not be neglected by designers of cryptographic devices. By evaluating the variability phenomena that occur at different spatial scales, we conclude that local effects are more damaging to the safety of countermeasures, due to the higher internal unbalancing it causes. Additionally, we verified that countermeasures implemented in smaller technologies present more damage to their protection, given the higher impact of local effects. Furthermore, the corner analysis, performed through deterministic simulations, showed to be insufficient to incorporate the variability factors, corroborating the need for statistical simulations. On the other hand, the influence of degradation due to the BTI effect did not manifest itself significantly in the safety of the countermeasures, showing a heterogeneous behavior, despite the negative impact observed on the delay of the circuits.