Sripath Mishra

Sripath Mishra

CS PhD candidate at Purdue. Anomaly and intrusion detection for edge and distributed systems — explainable AI, low-latency inference, federated learning, retrieval, GNNs. Seeking Applied Scientist / ML Engineer roles, 2027 start.

UAV-CAS: A Calibrated Digital-Twin Dataset for Intrusion Detection in UAV Swarm Networks

Published in arXiv:2606.17845 (dataset on IEEE DataPort), 2026

IDS models trained on wired-network benchmarks degrade sharply in real UAV swarms, and existing UAV-specific datasets do not systematically vary the mobility, link-quality, and routing conditions that cause the degradation — leaving no way to train or test against the shift that defeats them.

UAV-CAS comprises 99,492 flows from 1,024 configurations spanning five attack families (DoS, DDoS, blackhole, wormhole, replay) and nine collaborative compositions, generated by a Containernet digital twin. A four-layer calibration pipeline validates the twin against NSF AERPAW measurements across altitude-dependent path loss, mission-specific mobility, the link-level performance chain, and end-to-end trace fidelity: simulated-to-real RSS divergence is 0.33 Hellinger, below the 0.73 divergence measured between two independent real measurement campaigns.

Across ten baseline IDS, binary attack detection saturates above 0.98, confirming the data is learnable — but full attack-class identification remains hard, with per-class F1 from near zero to 0.82 and single digits for stealth attacks. Stealth-paired collaborative compositions fall to 0.51–0.86 AUROC.

Dataset, simulator, and calibration data are publicly released. → Dataset page

Authors: Sripath Mishra, Bharat Bhargava, Zizheng Liu, Shafkat Islam

Links

Recommended citation: Sripath Mishra, Bharat Bhargava, Zizheng Liu, Shafkat Islam. "UAV-CAS: A Calibrated Digital-Twin Dataset for Intrusion Detection in UAV Swarm Networks." arXiv:2606.17845, 2026. Dataset: IEEE DataPort, doi:10.21227/zgrg-z865.
Download Paper