Delving into adversarial attacks on deep policies

Abstract

Adversarial examples have been shown to exist for a variety of deep learning architectures. Deep reinforcement learning has shown promising results on training agent policies directly on raw inputs such as image pixels. In this paper we present a novel study into adversarial attacks on deep reinforcement learning polices. We compare the effectiveness of the attacks using adversarial examples vs. random noise. We present a novel method for reducing the number of times adversarial examples need to be injected for a successful attack, based on the value function. We further explore how re-training on random noise and FGSM perturbations affects the resilience against adversarial examples.

ICLR 2017 Workshop

References in corpus (1)

• Intriguing properties of neural networks

Cited by in corpus (47)

• Robust Physical-World Attacks on Deep Learning Models
• Adversarial Examples: Attacks and Defenses for Deep Learning
• Stochastic Activation Pruning for Robust Adversarial Defense
• Robust Deep Reinforcement Learning with Adversarial Attacks
• Adversarial Policies: Attacking Deep Reinforcement Learning
• Certified Adversarial Robustness for Deep Reinforcement Learning
• Robust Reinforcement Learning on State Observations with Learned Optimal Adversary
• Robust Deep Reinforcement Learning against Adversarial Perturbations on State Observations
• Whatever Does Not Kill Deep Reinforcement Learning, Makes It Stronger
• Detecting Adversarial Attacks on Neural Network Policies with Visual Foresight
• Adaptive Reward-Poisoning Attacks against Reinforcement Learning
• Defend Deep Neural Networks Against Adversarial Examples via Fixed and Dynamic Quantized Activation Functions
• Network Environment Design for Autonomous Cyberdefense
• Learning Visually-Grounded Semantics from Contrastive Adversarial Samples
• The Faults in Our Pi Stars: Security Issues and Open Challenges in Deep Reinforcement Learning
• Optimal Attacks on Reinforcement Learning Policies
• PAC-learning in the presence of evasion adversaries
• CopyCAT: Taking Control of Neural Policies with Constant Attacks
• Robust Deep Reinforcement Learning through Adversarial Loss
• Who Is the Strongest Enemy? Towards Optimal and Efficient Evasion Attacks in Deep RL
• Reinforcement Learning with Perturbed Rewards
• Falsification-Based Robust Adversarial Reinforcement Learning
• Provably Efficient Black-Box Action Poisoning Attacks Against Reinforcement Learning
• Robust Reinforcement Learning with Wasserstein Constraint
• Adaptive Adversarial Attack on Scene Text Recognition
• Blind Pre-Processing: A Robust Defense Method Against Adversarial Examples
• Optimal Attack and Defense for Reinforcement Learning
• Fooling Vision and Language Models Despite Localization and Attention Mechanism
• Understanding the Limits of Poisoning Attacks in Episodic Reinforcement Learning
• Better the Devil you Know: An Analysis of Evasion Attacks using Out-of-Distribution Adversarial Examples
• Network Defense is Not a Game
• Understanding Adversarial Attacks on Observations in Deep Reinforcement Learning
• One Bit Matters: Understanding Adversarial Examples as the Abuse of Redundancy
• Synthesis-guided Adversarial Scenario Generation for Gray-box Feedback Control Systems with Sensing Imperfections
• Reward Poisoning in Reinforcement Learning: Attacks Against Unknown Learners in Unknown Environments
• CROP: Certifying Robust Policies for Reinforcement Learning through Functional Smoothing
• Stealthy and Efficient Adversarial Attacks against Deep Reinforcement Learning
• Multi-Task Federated Reinforcement Learning with Adversaries
• Adaptive Adversarial Training for Meta Reinforcement Learning
• Certifiable Robustness to Adversarial State Uncertainty in Deep Reinforcement Learning
• Disturbing Reinforcement Learning Agents with Corrupted Rewards
• Controlled Caption Generation for Images Through Adversarial Attacks
• Augmenting Model Robustness with Transformation-Invariant Attacks
• Minimalistic Attacks: How Little it Takes to Fool a Deep Reinforcement Learning Policy
• Spatiotemporal Attacks for Embodied Agents
• Robust Reinforcement Learning under model misspecification
• Minimax Iterative Dynamic Game: Application to Nonlinear Robot Control Tasks