CILC addresses a threat from inside the team: a compromised robot can reconstruct approximations of an honest robot’s images and trajectory from the global descriptors it broadcasts for loop-closure detection. CILC applies secure multi-party computation only to the descriptor similarity step, so robots find inter-agent loop-closure candidates without revealing descriptors in the clear. Simulation and hardware tests show it stays real-time and within communication limits for visual and LiDAR descriptors.
Relevance
Encrypted radios stop outsiders; this paper takes the insider seriously, which is the robust robotics threat model. It also shows that the descriptors collaborative SLAM shares by default leak imagery and trajectory.
Links
- arXiv
- Subfield: Security, Multi-robot systems, Mapping and localisation