Hacking Predictors Means Hacking Cars: Using Sensitivity Analysis to Identify Trajectory Prediction Vulnerabilities for Autonomous Driving Security

arXiv:2401.10313v2 Announce Type: replace-cross
Abstract: Adversarial attacks on learning-based multi-modal trajectory predictors have already been demonstrated. However, there are still open questions about the effects of perturbations on inputs other than state histories, and how these attacks impact downstream planning and control. In this paper, we conduct a sensitivity analysis on two trajectory prediction models, Trajectron++ and AgentFormer. The analysis reveals that between all inputs, almost all of the perturbation sensitivities for both models lie only within the most recent …

abstract adversarial adversarial attacks analysis arxiv attacks autonomous autonomous driving cars control cs.cr cs.lg cs.ro cs.sy driving eess.sy effects hacking however identify impact inputs modal multi-modal planning prediction questions replace security sensitivity state trajectory type vulnerabilities