Cyber Resilience of Connected and Autonomous Transportation Systems (Phase II): Game-Theoretic Security Assurance of Network-Level Traffic Signal Control

Details

• Creators:
  Noruzoliaee, Mohamadhossein ; Hernandez, Mark ; Sauceda, Oziel ; Nazari, Fatemeh
• Corporate Creators:
  University of Texas Rio Grande Valley
• Corporate Contributors:
  Carnegie Mellon University. Traffic21 Institute. Safety21 University Transportation Center (UTC) ; United States. Department of Transportation. University Transportation Centers (UTC) Program ; United States. Department of Transportation. Office of the Assistant Secretary for Research and Technology ; United States. Department of Transportation. Federal Highway Administration
• Subject/TRT Terms:
  • Traffic signal controllers
  • Computer security
  • Connected vehicles
  • Autonomous vehicles
  • Flight recorders
  • Performance
  • Travel demand
  • Training
  • Risk assessment
  • Cybersecurity
  More +
• Publication/ Report Number:
  616
• Resource Type:
  Tech Report
• Geographical Coverage:
  United States
• Edition:
  Final Report (July 1, 2025 -July 31, 2026)
• Corporate Publisher:
  Carnegie Mellon University. Traffic21 Institute. Safety21 University Transportation Center (UTC)
• Description:
  Intelligent traffic signal control (TSC) is critical to connected and autonomous transportation systems, with important implications for mobility and safety. Its reliance on observed and communicated traffic information, however, creates cybersecurity vulnerabilities to adversarial information manipulation. This research investigates the robustness of network-level deep reinforcement learning-based TSC against strategic test-time state observation cyberattacks under blackbox and whitebox attacker settings. These attacks corrupt the signal controller's observations of the ground-truth traffic state, indirectly affecting traffic evolution through altered signal decisions. From a worst-case perspective, a strategic adversary selects bounded observation perturbations to maximize cumulative rather than purely instantaneous traffic degradation over the control horizon. The blackbox attacker operates without access to the victim signal control policy, while the whitebox attacker exploits policy access to construct policy-aware perturbations. Empirical cybersecurity assurance is pursued through adversarial training against learned attackers. Experiments in SUMO traffic simulator show that blackbox and whitebox attacks degrade the undefended signal controller's performance by approximately 50% and 73%, respectively. Adversarial training reduces these losses to nearly 6% and 7%, eliminating about 89% and 91% of attack-induced degradation with a moderate sacrifice in nominal performance and a favorable performance-robustness tradeoff. Sensitivity analyses reveal nonlinear effects of perturbation budget and dependencies on attacker information, traffic demand, and network size.
• Format:
  PDF
• Funding:
  69A3552344811 ; 69A3552348316
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/93556/dot_93556_DS1.pdf
• File Type:
  [PDF - 1.84 MB]
• Collection(s):
  Autonomous Vehicles ; University Transportation Centers
• Main Document Checksum:
  urn:sha-512:73a61ca712825066e5adfce58b7b373a70e9fd70ec973f4e44aa1c234c903f3035b360d493dffe420323be00951be665f65b36921dfa846d9e2ae88b590b5639