Quantum-Resilient UAV communications: performance and security tradeoffs for ML-KEM on ARM cortex-M4

This paper presents a simulation- and model-based study — all results are derived from validated simulation frameworks rather than physical hardware measurements — of the post-quantum migration of unmanned aerial vehicle (UAV) command-and-control links. The advent of cryptographically relevant quantum computers (CRQCs) fundamentally threatens the security of unmanned aerial systems (UAS), particularly command and control links that rely on classical public-key cryptography. Compounded by "harvest now, decrypt later” strategies, this necessitates a proactive transition to post-quantum cryptography (PQC). We develop a quantum threat model for UAV communications and evaluates a migration pathway anchored by the NIST-standardized ML-KEM (Kyber) algorithm. We model and benchmark ML-KEM on a simulated ARM Cortex-M4-class flight controller plat-form, analysing computational performance, memory footprint, energy consumption, and real-time scheduling behaviour. Results demonstrate that ML-KEM-768 encapsulation reduces latency by approximately 39% compared to elliptic curve Diffie-Hellman, while a hybrid ECDH+ML-KEM-768 configuration introduces moderate computational and memory overhead. However, the substantial increase in ciphertext size creates critical fragmentation bottlenecks on low-bandwidth telemetry links such as LoRa, motivating a proposed symmetric key derivation architecture for bandwidth-efficient rekeying. Side-channel analysis under an idealised leakage model confirms susceptibility to correlation power analysis, highlighting the necessity for hardened implementations. Furthermore, real-time operating system scheduling simulations under a two-task discrete-event model reveal that naive task co-scheduling induces significant jitter, necessitating strict priority isolation to maintain flight controller determinism. These findings establish simulation-grounded guidance for integrating PQC into resource-constrained UAV platforms without compromising real-time performance or communication reliability.