Abstract
Security attacks on intelligent transportation systems (ITS) may result in life-threatening situations. Combining deep neural networks with reinforcement learning (RL) models called DRL shows promising results when applied to urban Traffic Signal Control (TSC) for adaptive adjustment of traffic light schedules. In this paper, first, we explore the security vulnerabilities of DRL-based TSCs in the presence of adversarial attacks. We investigate the impact of the two distinct threat models with two state-of-the-art adversarial attacks using white-box and black-box settings. The attacks are simulated on different DRL-based TSC algorithms in a single intersection and multiple intersections. The results show that the performance of the DRL learning agent decreases in both adversarial attack models with white-box and black-box settings resulting in higher levels of traffic congestion. After analysing the adversarial attack models, we explored several sequential anomaly detection models. While sequential anomaly detection models minimizes the detection delays, it also achieves lower false alarm rates due to cumulative anomaly inspection. We also proposed an ensemble model that works with all the attack models without any model assumption. The results of anomaly detectors indicates that low-cost ensemble model achieves the best anomaly detection performance in all attack models and DRL settings.
Highlights
In recent years, data-driven approaches are often used to drive the design and performance evaluation of different control algorithms in Intelligent Transportation System (ITS)
Given the adversarial attacks Fast Gradient Sign Method (FGSM) and Jacobian-based Saliency Map Attack (JSMA) for single intersection and multi-intersection scenarios discussed in the previous subsection, we studied the performance of statistical anomaly detectors to detect even infinitesimally small anomalies
There are three main reasons why we employed a non-parametric sequential statistical anomaly detector for adversarial attacks on DRL-Traffic signal controller (TSC): (i) consecutive adversarial samples are more harmful for DRL controllers and need to be detected quickly, (ii) standard outlier detectors are susceptible to false alarms due to not considering temporal correlations in data, (iii) nonparametric sequential detectors have less miss-match error that results in lower detection error
Summary
Data-driven approaches are often used to drive the design and performance evaluation of different control algorithms in Intelligent Transportation System (ITS). Control loops like TSCs often use real-time traffic information (e.g., captured by local cameras/sensors or broadcast messages from vehicles) to perform intelligent control decisions. Learning based intelligent TSC agent collects messages from environment and schedules the traffic according to demand. Propose an online anomaly detection algorithm for detecting such adversarial attacks
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have