Search
Autonomous vehicles depend on cameras and sensors to identify road signs to ensure safe driving. However, the AI systems that underpin them are now confronting a novel form of threat known as 'visual attacks'. Researchers at the University of California, Santa Cruz, have made a groundbreaking discovery: attackers can insert specific textual cues into the environment to directly disrupt the decision-making processes of autonomous systems, such as self-driving cars and drones. This manipulation can lead these systems to engage in perilous behaviors. The study underscores the urgent need for the industry to develop new safety standards and protective measures promptly. The related research was published on the EurekAlert! website on the 27th and has been accepted for presentation at the 2026 IEEE Conference on Secure and Trustworthy Machine Learning.
-
C114 Communication Network -
Communication Home
