Detecting Perception-Based Attacks using Visual Odometry: Inconsistency Modeling and Checking on Robotic States
May 19, 2025·
,·
1 min read
Yuan Xu
Gelei Deng
Tianwei Zhang
Abstract
This work studies perception-based attacks against robotic systems and uses visual odometry to model and check inconsistencies in robotic states, helping detect attacks that manipulate a robot’s perception pipeline.
Type
Publication
2025 IEEE International Conference on Robotics and Automation (ICRA)
The paper continues my robotics-security work by connecting perception attack detection with state inconsistency modeling, using visual odometry as a practical signal for identifying compromised robotic perception.
