Environment perception technology is significant in the intelligent transportation system (ITS), which is the premise of the safety and stability of the unmanned vehicle system. The environmental perception system is mainly composed of lidar, camera, inertial measurement unit (IMU), global positioning system (GPS). Studies have shown that these sensors are vulnerable to external attacks, resulting in the distortion of the information perceived by the unmanned vehicle, and ultimately resulting in wrong driving decisions, which bring serious safety threats to life and property. In this paper, we utilize correlation of sensors in the space domain to establish distance models between multi-sensor and the distance models of a single sensor in time domain respectively. With the established distance models and distance distributions, we can accurately identify anomalous or attacked sensors with a certain degree of confidence. The hierarchical detection method of spatial correlation and temporal correlation is adopted, which makes that this method not only takes into account the detection efficiency, but also ensures the detection accuracy. Our experimental results quantitatively show that the method realizes real-time attack detection, and prove the effectiveness and robustness of the proposed method based on the open source KITTI datasets.