AI & Technology · Doctoral Studies
Preventing Adversarial Attacks on Autonomous Driving Models: Threats, Defenses, and a 95% Accuracy Breakthrough Autonomous driving systems are among the most exceptional technological developments of recent times — but they are also vulnerable to a sophisticated class of threat known as adversarial attacks. These carefully crafted perturbations can cause a self-driving vehicle to misread a stop sign, ignore a pedestrian, or execute a dangerous maneuver, often with changes invisible to the human eye.
This research paper investigates the behavior of end-to-end autonomous driving models when confronted with physical adversarial attacks. Five attack methods were applied to three models, four defense strategies were evaluated, and the results achieved a classification accuracy of 95% — demonstrating that robust, layered defenses can significantly harden autonomous driving AI against real-world threats.
Jul 28, 2026 
