AH-KEM: Adaptive Hybrid Key Encapsulation Mechanism for Post-Quantum Secure Cloud Communication

Abstract

Classical public-key cryptography (RSA-2048, ECDH) is vulnerable to Shor's algorithm on quantum computers, with harvest-now-decrypt-later threats already targeting archived cloud traffic for future decryption. While NIST's 2024 ML-KEM (FIPS 203) and ML-DSA (FIPS 204) standards provide quantumresistant primitives, uniform PQC deployment across diverse cloud workloads remains computationally inefficient and lacks validated hybrid security and scalability. This paper introduces AH-KEM, an adaptive hybrid key encapsulation mechanism for TLS 1.3-compatible cloud communication. The AH-KEM algorithm determines a risk rating for each session from three dimensions-sensitivity of the data, expected length of the session, and network exposure-and then uses a deterministic method to choose among four modes of cryptography, which include Classical-only (X25519/ECDSA), Parallel-Hybrid (X25519/ML-KEM-512/ML-DSA-44), PQC-Dominant(ML-KEM-768/ML-DSA-65), and Full-Hybrid (X25519/ML-KEM-1024/ML-DSA-87). The safe dual-PRF combiner combines the classical and post-quantum shared keys to obtain a single session key that is secure in the Quantum Random Oracle Model if either KEM used is secure. Implemented in Python with FIPS-compliant software stubs on a Lenovo LOQ 15APH8 (AMD Ryzen 7 7840HS), AH-KEM demonstrates that, when both are generated fresh per session to satisfy forward secrecy, ML-KEM-768 key generation is over 11, 000× faster than RSA-2048 (0.004 ms vs. 44.335 ms); note RSA keys are not conventionally regenerated per session in deployed TLS, so this comparison isolates the cost of per-session ephemeral generation rather than RSA's typical amortized cost. Full-Hybrid end-to-end handshake latency is 0.514 ms (a sub-millisecond overhead); servers sustain zero errors at 250 concurrent clients; and risk classification achieves 100.0% accuracy over 10,000 sessions.