Event

Research on Sybil Attack Defense for Multi-UAV Exploration

1 reports · 1 sources

Shyaam Mahesh and other authors submitted a preprint paper studying the problem of Sybil attacks when multiple UAVs collaboratively explore buildings: a compromised UAV forges identities, seizes frontier positions, and sends false scans, causing real UAVs to miss areas and generating nonexistent walls in the map. The authors propose five defense methods that rely only on the UAVs' own sensors and maps, including map data auditing, checking the physical presence of a body at an identity's location, trusting only identities of bodies that have been observed, isolating identities not observed by any UAV, and binding observed bodies to a single identity. Simulation tests cover seven categories totaling 70 maps, including offices, tunnels, a three-story building, and caves. The authors report that across more than 2,350 missions, honest UAVs revoked trust in teammates only twice; exceptions remain, such as two UAVs colluding, a forged body adjacent to a real UAV, 3 of 10 caves, and a formation of 16 UAVs, with the latter three fixable by "binding bodies to identities."