LiDAR point cloud transmission: Adversarial perspectives of spoofing attacks in autonomous driving

Abstract

LiDAR technology uses laser light to illuminate the surrounding area and detect 3D objects. Calculates different features such as distance, shape, height, and direction of objects, ultimately generating comprehensive 3D maps by collecting cloud points. They are frequently used in autonomous vehicles, robotics, forestry, archaeology, and environmental monitoring. LiDAR is important in autonomous vehicles for recognizing objects, pedestrians, and other vehicles, allowing them to make judgments to prevent collisions and ensure human safety. The LiDAR systems are generally robust; they are not immune to certain types of security attacks that could compromise the integrity of the signals and may affect the accuracy of the data. If the signal is compromised, the system could incorrectly interpret the environment, resulting in erroneous object recognition, incorrect obstacle avoidance decisions, or inaccurate environment mapping. As a result, it can lead to serious consequences, such as property damage, accidents, or dangerous driving conditions. To address these security challenges and establish better security mechanisms for LiDAR systems, we have proposed a novel technique for detecting and avoiding all possible spoofing attacks on LiDAR signals. Initially, the system identifies potential spoofing attacks, and as a preventive measure, it employs an optimized path strategy. This strategy ensures safe crossings and autonomous navigation while avoiding obstacles along the vehicle’s route. The main aim is to identify the spoofed objects, suitably map the 3D presentation of the objects, and properly navigate autonomous vehicles with an optimized path selection in the automatic driving system. The proposed system is validated in different scenarios, and the experimental results demonstrate a success rate of 94.57% in true positive and false positive rates, indicating the effectiveness of the system. The average precision rate of 0.95 further supports its performance. The strength of the system was confirmed by testing it with different intersection over union (IoU) rates in different situations and closely looking at the attacker’s success rate.