Data-Driven Modeling and Anomaly Detection of Attacked HATS

Details

• Creators:
  Homaifar, Abdollah
• Corporate Creators:
  North Carolina Agricultural and Technical State University ; Center for Automated Vehicles Research with Multimodal Assured Navigation (CARMEN+) Tier-1 University Transportation Center (UTC)
• Corporate Publisher:
  Center for Automated Vehicles Research with Multimodal Assured Navigation (CARMEN+) Tier-1 University Transportation Center (UTC)
• Corporate Contributors:
  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:
  • Automated vehicle control
  • Transportation safety
  • Computer security
  • Global Positioning System
  • Machine learning
  • Validation
  • Visual media
  • Navigation systems
  • Time series analysis
  • Disaster resilience
  More +
• Resource Type:
  Tech Report
• Geographical Coverage:
  United States
• Edition:
  Final Report: (Nov. 1, 2023 to August30, 2026)
• Description:
  This project developed data-driven methods to improve the cybersecurity, verification and validation (V&V), and resilience of Highly Automated Transportation Systems (HATS) under cyberattack. Subtask 1 focused on Global Navigation Satellite System (GNSS) spoofing detection using onboard vehicle measurements, adaptive DBSCAN, and time-series analysis, and extended the work toward stereo visual navigation when GNSS becomes untrustworthy. The data-driven detector reduced the nominal displacement threshold from 3.35 m to 1.79 m; adaptive DBSCAN detected below-threshold multiple small biased attacks with 98.38% accuracy; and the time-series framework detected 0.28 m spoofing steps with 93.66% accuracy. The ongoing visual-navigation framework achieved 88.50% and 88.00% direction-classification accuracy for the two global displacement axes and a 7.12 mm translational relative-pose root-mean-square error. Subtask 2 developed a fuzzy-equivalence V&V framework that produced a 67.24% Adjusted Accuracy and 89.57% estimated peak accuracy benchmark, and a GNSS onset-localization framework whose enhanced neural model achieved zero-sample median error and 4.75-sample mean absolute error at a sampling rate of 10 Hz.
• Format:
  PDF
• Funding:
  69A3552348327
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/94254/dot_94254_DS1.pdf
• File Type:
  [PDF - 3.44 MB]
• Collection(s):
  University Transportation Centers
• Main Document Checksum:
  urn:sha-512:985bc7377298d7baab3f1dab0110ad8c94416a93fe6fd5e7e2cc2bbc8daf31afd5b367f96e3215d742cffebea822a14799d444477b8b2ee3f17f2580d291b870