Mitigating Deep Reinforcement Learning Backdoors in the Neural Activation Space
This addresses a security threat for DRL systems by improving detection of hidden backdoors, though it is incremental as it builds on existing defense concepts.
The paper tackles the problem of detecting elusive in-distribution backdoor triggers in Deep Reinforcement Learning agents by proposing a method that analyzes neural activation patterns, showing that lightweight classifiers can prevent malicious actions with considerable accuracy.
This paper investigates the threat of backdoors in Deep Reinforcement Learning (DRL) agent policies and proposes a novel method for their detection at runtime. Our study focuses on elusive in-distribution backdoor triggers. Such triggers are designed to induce a deviation in the behaviour of a backdoored agent while blending into the expected data distribution to evade detection. Through experiments conducted in the Atari Breakout environment, we demonstrate the limitations of current sanitisation methods when faced with such triggers and investigate why they present a challenging defence problem. We then evaluate the hypothesis that backdoor triggers might be easier to detect in the neural activation space of the DRL agent's policy network. Our statistical analysis shows that indeed the activation patterns in the agent's policy network are distinct in the presence of a trigger, regardless of how well the trigger is concealed in the environment. Based on this, we propose a new defence approach that uses a classifier trained on clean environment samples and detects abnormal activations. Our results show that even lightweight classifiers can effectively prevent malicious actions with considerable accuracy, indicating the potential of this research direction even against sophisticated adversaries.