Secure and Robust Machine Learning for Autonomous Driving Systems

As autonomous driving systems (ADS) become increasingly prevalent in modern transportation, critical concerns have emerged regarding their security vulnerabilities and performance inconsistency, particularly in pedestrian detection and natural language processing components. Current machine learning technologies, while effective, can introduce variability where the model performance remains uniform across different scenarios and is susceptible to security attacks that may compromise both safety and robustness. This research aims to enhance the security and robustness of autonomous driving systems through a comprehensive investigation of vulnerabilities and the development of novel protective strategies. The project will focus on identifying and analyzing security and robustness vulnerabilities in ADS, developing novel strategies to promote robustness and enhance security in pedestrian detection systems, improving robustness in automotive large language models, and implementing prototype systems for real-world evaluation. The research methodology encompasses four integrated tasks. First, the research team will develop a novel consistency poisoning attack framework to assess system vulnerabilities. Second, the research team will analyze and mitigate consistency vulnerabilities in pedestrian detection systems through advanced machine-learning techniques. Third, the research team will enhance consistency in Large Language Models through innovative prompt engineering and model fine-tuning. Finally, the research team will implement prototype systems and conduct comprehensive evaluations using real-world datasets to validate its th approaches. This project directly aligns with the key priorities of the U.S. Department of Transportation (USDOT). First, it supports the Safety strategic goal by developing robust defenses against security and consistency attacks in autonomous vehicles, particularly focusing on protecting vulnerable road users through enhanced pedestrian detection systems. Secondly, the research advances the Department's Cybersecurity priority by addressing AI vulnerability in autonomous systems that could impact various communities. The outcomes of robust and consistent machine learning directly support the USDOT commitment to promoting transportation safety and ensuring that emerging technologies benefit all Americans. In summary, the project's comprehensive approach to addressing both technical and societal challenges in autonomous driving systems demonstrates strong alignment with the USDOT vision for safe and sustainable transportation innovation.

• Record URL:
  • https://www.clemson.edu/cecas/tracr/research/projects/#secureandrobustmachinelearningforautonomousdrivingsystems

Language

• English

Project

• Status: Active
• Funding: $353,945.00
• Contract Numbers:

  69A3552344812

• Sponsor Organizations:

  Office of the Assistant Secretary for Research and Technology

  University Transportation Centers Program
  Department of Transportation
  Washington, DC  United States  20590

  Clemson University

  216 Lowry Hall
  Clemson, SC, SC  United States  29634

  University of Texas at Dallas

  800 W Campbell Rd
  Richardson, Texas  United States  75080
• Managing Organizations:

  National Center for Transportation Cybersecurity and Resiliency (TraCR)

  Clemson University
  Clemson, SC  United States 

  Clemson University

  216 Lowry Hall
  Clemson, SC, SC  United States  29634
• Project Managers:

  Chowdhury, Mashrur

• Performing Organizations:

  Clemson University

  216 Lowry Hall
  Clemson, SC, SC  United States  29634

  University of Texas at Dallas

  800 W Campbell Rd
  Richardson, Texas  United States  75080
• Principal Investigators:

  Wu , Yongkai

  Luo, Feng

  Khan, Latifur

  Chowdhury, Mashrur

  Thuraisingham, Bhavani

• Start Date: 20250101
• Expected Completion Date: 20251231
• Actual Completion Date: 0
• USDOT Program: University Transportation Centers

Subject/Index Terms

• TRT Terms: Autonomous vehicles; Computer security; Information processing; Machine learning; Pedestrian detection; Traffic safety
• Subject Areas: Data and Information Technology; Highways; Pedestrians and Bicyclists; Safety and Human Factors; Security and Emergencies; Vehicles and Equipment;

Filing Info

• Accession Number: 01950197
• Record Type: Research project
• Source Agency: National Center for Transportation Cybersecurity and Resiliency (TraCR)
• Contract Numbers: 69A3552344812
• Files: UTC, RIP
• Created Date: Mar 27 2025 3:28PM