Certified Adversarial Robustness for Deep Reinforcement Learning

Björn Lütjens, Michael Everett, Jonathan P. How
Proceedings of the Conference on Robot Learning, PMLR 100:1328-1337, 2020.

Abstract

Deep Neural Network-based systems are now the state-of-the-art in many robotics tasks, but their application in safety-critical domains remains dangerous without formal guarantees on network robustness. Small perturbations to sensor inputs (from noise or adversarial examples) are often enough to change network-based decisions, which was already shown to cause an autonomous vehicle to swerve into oncoming traffic. In light of these dangers, numerous algorithms have been developed as defensive mechanisms from these adversarial inputs, some of which provide formal robustness guarantees or certificates. This work leverages research on certified adversarial robustness to develop an online certified defense for deep reinforcement learning algorithms. The proposed defense computes guaranteed lower bounds on state-action values during execution to identify and choose the optimal action under a worst-case deviation in input space due to possible adversaries or noise. The approach is demonstrated on a Deep Q-Network policy and is shown to increase robustness to noise and adversaries in pedestrian collision avoidance scenarios and a classic control task.

Cite this Paper

BibTeX
@InProceedings{pmlr-v100-lutjens20a, title = {Certified Adversarial Robustness for Deep Reinforcement Learning}, author = {L\"utjens, Bj\"orn and Everett, Michael and How, Jonathan P.}, booktitle = {Proceedings of the Conference on Robot Learning}, pages = {1328--1337}, year = {2020}, editor = {Kaelbling, Leslie Pack and Kragic, Danica and Sugiura, Komei}, volume = {100}, series = {Proceedings of Machine Learning Research}, month = {30 Oct--01 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v100/lutjens20a/lutjens20a.pdf}, url = {https://proceedings.mlr.press/v100/lutjens20a.html}, abstract = {Deep Neural Network-based systems are now the state-of-the-art in many robotics tasks, but their application in safety-critical domains remains dangerous without formal guarantees on network robustness. Small perturbations to sensor inputs (from noise or adversarial examples) are often enough to change network-based decisions, which was already shown to cause an autonomous vehicle to swerve into oncoming traffic. In light of these dangers, numerous algorithms have been developed as defensive mechanisms from these adversarial inputs, some of which provide formal robustness guarantees or certificates. This work leverages research on certified adversarial robustness to develop an online certified defense for deep reinforcement learning algorithms. The proposed defense computes guaranteed lower bounds on state-action values during execution to identify and choose the optimal action under a worst-case deviation in input space due to possible adversaries or noise. The approach is demonstrated on a Deep Q-Network policy and is shown to increase robustness to noise and adversaries in pedestrian collision avoidance scenarios and a classic control task.} }
Endnote
%0 Conference Paper %T Certified Adversarial Robustness for Deep Reinforcement Learning %A Björn Lütjens %A Michael Everett %A Jonathan P. How %B Proceedings of the Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2020 %E Leslie Pack Kaelbling %E Danica Kragic %E Komei Sugiura %F pmlr-v100-lutjens20a %I PMLR %P 1328--1337 %U https://proceedings.mlr.press/v100/lutjens20a.html %V 100 %X Deep Neural Network-based systems are now the state-of-the-art in many robotics tasks, but their application in safety-critical domains remains dangerous without formal guarantees on network robustness. Small perturbations to sensor inputs (from noise or adversarial examples) are often enough to change network-based decisions, which was already shown to cause an autonomous vehicle to swerve into oncoming traffic. In light of these dangers, numerous algorithms have been developed as defensive mechanisms from these adversarial inputs, some of which provide formal robustness guarantees or certificates. This work leverages research on certified adversarial robustness to develop an online certified defense for deep reinforcement learning algorithms. The proposed defense computes guaranteed lower bounds on state-action values during execution to identify and choose the optimal action under a worst-case deviation in input space due to possible adversaries or noise. The approach is demonstrated on a Deep Q-Network policy and is shown to increase robustness to noise and adversaries in pedestrian collision avoidance scenarios and a classic control task.
APA
Lütjens, B., Everett, M. & How, J.P.. (2020). Certified Adversarial Robustness for Deep Reinforcement Learning. Proceedings of the Conference on Robot Learning, in Proceedings of Machine Learning Research 100:1328-1337 Available from https://proceedings.mlr.press/v100/lutjens20a.html.

Related Material