Adversarial Attacks on Multi-task Visual Perception for Autonomous Driving

In recent years, deep neural networks (DNNs) have accomplished impressive success in various applications, including autonomous driving perception tasks. However, current are easily deceived by adversarial attacks. This vulnerability raises significant concerns, particularly safety-critical applications. As a result, research into attacking and defending DNNs has gained much coverage. this work, detailed attacks applied on diverse multi-task visual network across distance estimation, semantic segmentation, motion detection, object detection. The experiments consider both white black box for targeted un-targeted cases, while task inspecting the effect all others, addition to of applying simple defense method. We conclude paper comparing discussing experimental results, proposing insights future work. visualizations available at <uri>https://youtu.be/6AixN90budY</uri>.