A Crypto-Agile Post-Quantum Architecture for Banking Systems

Post-quantum cryptography is usually evaluated one algorithm at a time: can ML-KEM encrypt a session, can ML-DSA sign a transaction. A bank is not one application, though, but an ecosystem of customer channels, ATMs, point-of-sale terminals, USSD services, core banking systems, internal PKI and hardware security modules, and interbank messaging, each with different constraints and each currently secured by classical public-key cryptography that a cryptographically relevant quantum computer would break. This paper asks a narrower and more concrete question than whether post-quantum cryptography works: where, specifically, it belongs inside that ecosystem, and how it should be built in so the answer does not depend permanently on whichever algorithms are standardized today. It proposes a three-layer security model (communication, transaction, and long-term cryptographic infrastructure) built around a crypto-agile cryptographic-service abstraction rather than direct application-level dependence on any single algorithm, evaluates representative post-quantum mechanisms from different mathematical families (ML-KEM and HQC for key establishment; ML-DSA, SLH-DSA, and FN-DSA for signatures) against their documented security, performance, and size characteristics for different banking components, and develops a conceptual, component-differentiated migration path from classical through hybrid to crypto-agile post-quantum infrastructure. The design is literature-based and conceptual, so the paper does not build a complete banking system; it draws on finalized and draft standards, published benchmarks, and existing banking-security and crypto-agility research, and states explicitly what it does not claim. It is a companion to, and assumes without re-arguing, the case this research team's earlier paper made for why African financial institutions need to prepare for post-quantum migration now.