Post-quantum Federated Learning: Secure And Scalable Threat Intelligence For Collaborative Cyber Defense

This is a 29-day-old research paper (not a deployed project) with zero GitHub adoption signals. The core contribution—applying NIST-standardized PQC algorithms to federated learning for threat intelligence—is a technically sound novel combination of existing, well-understood components (CRYSTALS-Kyber for KEM, CRYSTALS-Dilithium for signatures, federated averaging). However, it lacks production implementation, real-world deployment, user adoption, or comparative benchmarks against classical FL on realistic threat-intel datasets. The paper appears to be a proof-of-concept demonstrating vulnerability of RSA-encrypted gradients to simulated quantum attacks and proposing PQC alternatives. Frontier labs (Google, Meta, OpenAI) have active quantum-safe cryptography and federated learning programs; they could trivially integrate NIST PQC into their FL frameworks. The threat model (harvest-now-decrypt-later attacks on FL) is real but the solution is straightforward application of published standards. The project carries high frontier risk because: (1) PQC is part of major cloud platforms' roadmaps; (2) FL infrastructure is already being hardened by major labs; (3) integrating Kyber/Dilithium into gradient encryption is not a defensible differentiation. A deployed, production FL system with threat-intel use cases, real organizational adoption, and benchmark superiority would score 5-6; a reference implementation with zero deployment scores 1-3. This sits at 2 because the novelty is real but the execution is purely academic without ecosystem traction.