SPADE: A Protocol for Secure Post-Quantum AI-Model Deployment to Edge-Devices
Soumyadev Saha, Ranajoy Hajra, Shashidhara R +3 more
The deployment of over-the-air model updates in edge AI systems is hindered by suboptimal distribution protocols and a dependence on classical cryptographic methods, which are susceptible to potential threats from quantum cryptanalysis in the future. To address this vulnerability, we propose the Secure Post-quantum AI-model Deployment to Edge devices (SPADE) protocol. SPADE protects the complete model update lifecycle by combining NIST-standardized ML-KEM for post-quantum key exchange with EdDSA for deterministic digital signatures, ensuring end-to-end confidentiality, integrity, and authenticity. We formalize the security of SPADE against both classical and quantum adversaries through rigorous threat modeling and standard game-based security proofs. Furthermore, we implement and benchmark SPADE against three classical and decentralized baseline protocols (including ECDH-ChaCha20 and IPFS-AES) on both server and mobile hardware platforms. Our empirical evaluation demonstrates that while the post-quantum cryptographic primitives introduce an anticipated computational overhead, the latency scales linearly and remains highly competitive with vulnerable classical methods. SPADE thereby delivers a practical, forward-secrecy-preserving framework for securing modern edge AI deployments.
IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT)•Jul 1, 2026•0 citations