Research On Security Defense For Ultrasonic Sensor Attack

Sensors are an important measurement device that autonomous vehicles rely on,and their safety and reliability play a crucial role in autonomous driving decisions.Among the sensors equipped in autonomous vehicles,ultrasonic sensors are mainly used for obstacle detection at close range.Unfortunately,at the 2016 DEFCON conference,the researchers used only a tiny ultrasonic transmitter to easily interfere with the Tesla car's ultrasonic ranging device,causing the Tesla car's ranging model to fail and emergency braking.In response to this security issue,many scholars began to conduct research on ultrasonic attack detection.The most classic work is the PSA attack detection algorithm proposed by Xu et al in 2018,which can detect the characteristics of echo frequency,phase,amplitude,duration,etc.,but this method ignores the complex motion conditions in the dynamic environment and needs to be increased.Third-party more expensive equipment,which may also introduce security problems caused by new equipment.In view of its shortcomings,the latest improvement is the attack detection model proposed by Lee et al.in 2019,which takes into account the impact of the Doppler effect on the attack detection model under motion,while also controlling the relatively low detection cost.However,when the distance between the sensor and the obstacle exceeds 1.5 meters,or the sensor beam angle is greater than 50 degrees,the detection success rate may decrease sharply,which may lead to certain problems in the promotion and use of this model..This thesis proposes a multi-featured ultrasonic attack detection method for the problem of lower detection success rate at larger distances and amplitude angles.It mainly encodes the uncertain period and duration dynamically,and designs a sliding-based method that fits the dynamic environment changes.The linear detection model of the window and the upper and lower limit detection model based on the second-order linear invariant function,through the cooperative detection of the two,the matching verification of the ultrasonic echo information and the system coding information from multiple features is fully considered,and the motion state is fully considered The effect of Doppler effect on the ultrasonic attack detection model has realized the matching algorithm based on the echo duration and the predetermined state of the system.Under the premise of effective coding cost,the detection accuracy is improved,and the detection time is reduced.Experiments have been carried out on AJ-SR04 M ultrasonic sensor and HC-SR04 ultrasonic sensor,which are more commonly used in the current market,and the results show that the method in this thesis can still effectively detect ultrasonic attacks in the range of 0 to 6 meters at a beam angle of 75 degrees.The maximum detection time does not exceed 0.6 seconds,which has relatively good practicality.