RobustUAVs.ai
An end-to-end UAV security benchmark and a composed network-to-navigation
certificate. The interactive client needs JavaScript; the substance is below.
What this is
UAV security research is split in two. Network-layer work studies the swarm
mesh, the command-and-control link and the intra-vehicle bus. Autonomy-layer
work studies GNSS spoofing, controller robustness and mission completion.
Neither half tells you what an attacker who defeats the first can do to the
second. This project closes that gap with a unified six-source benchmark and
a composition theorem carrying a detector operating point through to a
certified Mission-Completion-Rate floor.
The six sources
- UAV-EW-Bench — autonomy, mission and GNSS, 93,600 flights
- UAVIDS-2025 — network, swarm mesh, 122,171 flows
- DATAMUt — network, swarm mesh, per-hop traces
- HCRL UAVCAN — network, intra-vehicle bus, 10 scenarios
- UAV Attack Dataset — autonomy, GNSS, 3 live PX4 flights
- UAV-CAS — network, swarm mesh, flow statistics
Artifact
The schema, per-dataset adapters, certificate engine, experiment scripts and
every committed result are browsable at /artifact/.
All six datasets are published on Kaggle under
DOI 10.34740/kaggle/dsv/18346203.
Roger Nick Anaedevha, Institute of Cyber Intelligent Systems, MEPhI ·
Keiwan Soltani, Missouri S&T · Federico Corò, University of Padova.