AI-driven adaptive adversaries and the erosion of cryptographic trust in public key systems

Abstract This paper examines the erosion of Public Key Cryptography (PKC) security under adaptive adversarial optimisation driven by artificial intelligence. The problem addressed is the growing mismatch between algorithm-centric cryptographic security models and operational attack realities, where adversaries exploit implementation-level observability rather than breaking cryptographic primitives. The methodology integrates a reproducible bibliometric analysis of Web of Science records, qualitative evidence from twenty expert interviews and three industry workshops, and a technical synthesis of AI-enabled attack mechanisms across the cryptographic lifecycle. Results show that existing research is structurally concentrated on algorithmic robustness, with no significant focus on AI-driven attack vectors, while 82% of practitioners attribute private key compromise to AI-augmented optimisation and side-channel inference. The paper’s contribution is fourfold: (1) identification of a systemic research gap in AI-enabled cryptographic attacks; (2) development of an adaptive adversarial threat model spanning key generation to validation; (3) empirical validation of implementation-layer compromise mechanisms; and (4) formulation of AI-aware cryptographic resilience requirements extending beyond post-quantum approaches. The findings demonstrate that cryptographic security must be reconceptualised as an adaptive, system-level property rather than a function of algorithm strength alone.