Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore a groundbreaking security research presentation that investigates vulnerabilities in autonomous vehicle perception systems through luminescent adversarial attacks on lane markers. Discover how researchers from the University of Texas at Arlington and Clemson University demonstrate novel attack methods that exploit the textural properties of luminescent lane markers to deceive deep neural network-based lane detection models used in autonomous vehicles. Learn about the comprehensive experimental methodology that includes both digital simulations and physical testing using Openpilot devices, revealing complete model failure in worst-case scenarios and approximately 33% failure rates in optimal conditions. Understand the significant safety implications of these attacks during nighttime and low-light driving conditions, where autonomous driving systems rely heavily on accurate lane marker detection for safe navigation. Gain insights into the systematic exposure of vulnerabilities in state-of-the-art lane detection models and the critical need for robust defensive measures to protect autonomous vehicles from these emerging adversarial threats that could lead to serious safety violations.
