SKYTRA: Traceable Key Encapsulation for Secure Multi-Agent Communications in Embodied UAV Networks
Ming Xie, Mingfei Zeng, Zhede Gu +4 more
Agentic AI, particularly in the form of embodied agents such as unmanned aerial vehicle (UAV) swarms, requires secure and efficient communication to enable collaborative decision-making. However, the open wireless links these agents use for coordination are vulnerable to eavesdropping and adversarial attacks. Because each command capsule must reach many recipients simultaneously, classical per-recipient public-key encryption inflates airtime, drains batteries, and complicates key management. Moreover, a captured drone can leak its secret key, so the compromised unit must be identifiable without grounding the entire fleet, and the cryptosystem must remain secure against future quantum adversaries. In this paper, we propose a traceable hybrid multi-recipient key encapsulation mechanism (mKEM) to meet these requirements, called SKYTRA-KEM. Our scheme enables secure group-oriented communication for multi-agent systems, where a single compact ciphertext serves all swarm members that share a group public key while holding distinct, traceable secret keys. The experiment shows that, for a swarm of 100 drones arranged in five groups, the scheme reduces sender CPU time from 1.31 ms to 0.20 ms and cuts broadcast size from 78.1 kB to 4.0 kB. A qualitative comparison with recent mKEM constructions further demonstrates the unique combination of hybrid security, traceability, and bandwidth efficiency provided by SKYTRA-KEM. These results demonstrate an accountable, quantum-resilient, and energy-efficient communication framework designed for the resource-constrained nature of embodied agents.
IEEE Transactions on Cognitive Communications and Networking•Jan 1, 2026•0 citations