Design And Implementation Of GNSS Anti-Spoofing Algorithm For Autonomous Vehicle

The rapid development of automatic driving technology not only brings convenience and efficiency,but also faces more potential risks.Driving safety is one of the core issues.As an important part of autonomous vehicle,GNSS can deceive driving decisions and even jeopardize traffic safety when the vehicle is deceived.Therefore,preventing from GNSS deception is also an indispensable work to ensure driving safety.Starting from the working mechanism of autonomous vehicle,this paper studies the methodology of detecting GNSS spoofing attacks through sensor data.The detection scheme based on vehicle sensors proposed in this paper can detect real-time GNSS spoofing attacks without adding additional devices by automatically driving sensors on the vehicle itself.According to the two specific spoofing attack scenarios analyzed,two targeted anti-spoofing methods are proposed.One is to calculate the position deviation and judge whether it exceeds the threshold through the prediction of the position.The second is to judge the consistency of steering through steering detection and combining with the motion state.The main contributions of this paper are as follows:1.This paper analyzes the possible GNSS spoofing attack scenarios,with the consideration of latency and cost,and proposes a GNSS spoofing detection scheme based on sensor data.Vehicle borne sensors are cheap and highly secured.They transmit information through CAN,born high data acquisition frequency and are closer to the driving state of autopilot compared with GNSS.Therefore,real-time detection of spoofing attacks can be realized.2.This paper proposes a GNSS anti spoofing algorithm based on CNN-LSTM.After feature selection,the model uses IMU and can data and current position data to predict the next position.Then calculate the distance between the predicted position and the GPS signal position obtained by the GPS receiver.Finally,compare it with the determined distance threshold.3.This paper presents a GNSS anti spoofing algorithm based on KNN-DTW.The vehicle steering angle data obtained by the sensor is used to detect the vehicle steering action.Combine the detected steering category results with the steering displayed by the navigation system and the motion state detected by speed to detect whether they are consistent.This paper combines the real autopilot dataset comma2k19 to generate supervised learning data,and simulates the detection algorithm.In the GNSS anti spoofing algorithm based on CNN-LSTM,the model with feature selection has better prediction effect.In the GNSS anti spoofing algorithm based on KNN-DTW,all turns are correctly judged.At the same time,the spoofing attack data is constructed to verify the algorithm.The results show that the proposed algorithm is effective in detecting GNSS spoofing attacks.