The environmental perception system is a vital part of unmanned driving technology, and is a prerequisite for the safety and stability of the entire unmanned vehicle. At present, the research on environment perception technology in the field of unmanned driving mainly focuses on the acquisition of environmental information in ideal environments, high-precision semantic information recognition, and multi-sensor information fusion, etc., without forming a comprehensive attack detection and defense system. This project studies real-time attack defense and high-precision data recovery methods for unmanned system perception equipment. By studying the overlap relationship and semantic relevance of information between different sensors, the mathematical model of information crossing and the mathematical model of false information interference are established, the fault-causing sensor is judged according to the information crossing model, and the distortion information is reproduced with high precision by methods such as matrix completion. Constitution recovery. With the development of the unmanned vehicle industry, unmanned vehicle attacks and security research will highlight its importance. This research is of great practical significance in scientific research and engineering applications.