The shared codes and data include necessary materials for reproducing the results of our paper "Stop-and-Go: Exploring Backdoor Attacks on Deep Reinforcement Learning-based Traffic Congestion Control Systems".

Trigger_exploration: trigger_exploration_01.m: trigger exploration for the congestion attack and the insurance attack

Neural networks: Containing the benign and backdoor models for single-lane ring (single_lane_notrigger.sav, single_lane_congestion.sav, single_lane_insurance.sav)

In our paper, the training of benign models are based on the DDPG algorithm and thanks to the ddpg agent building by Chris Yoon on GitHub.