Camdar-Adv - Method for Generating Adversarial Patches on 3D Object.

DNN model is the core technology for sensors of the autonomous driving platform to perceive the external environment. However, it has a certain vulnerability, and the artificial designed adversarial examples can make the DNN model output the wrong results. These adversarial examples not only exist in the digital world, but also in the physical world. At present, researches on autonomous driving platform mainly focus on attacking a single sensor. In this paper, we presnet a method called Camdar-adv for generating adversarial examples, which can attack the optical image sensor based on any 3D object. Specifically, based on a 3D object that can attack LiDAR sensors, a geometric transformation can be used to project it onto the 2D plane. Perturbation can be added on the 2D plane to generate 2D adversarial examples, which can attack the optical image sensor in the black-box setting, without changing the object’s geometry.