Computer Engineering Commons^™

Adversarial Patch Attacks On Deep Reinforcement Learning Algorithms, Peizhen Tong

McKelvey School of Engineering Theses & Dissertations

Adversarial patch attack has demonstrated that it can cause the misclassification of deep neural networks to the target label when the size of patch is relatively small to the size of input image; however, the effectiveness of adversarial patch attack has never been experimented on deep reinforcement learning algorithms. We design algorithms to generate adversarial patches to attack two types of deep reinforcement learning algorithms, including deep Q-networks (DQN) and proximal policy optimization (PPO). Our algorithms of generating adversarial patch consist of two parts: choosing attack position and training adversarial patch on that position. Under the same bound of total …