Purdue researchers uncover vulnerability in autonomous vehicle vision systems
09-04-2025
Raymond Muller, a recent PhD graduate from Purdue’s Department of Computer Science.
Self-driving cars rely on advanced perception systems to make split-second decisions—detecting pedestrians, reading traffic lights, and avoiding obstacles. But what if those systems are overwhelmed? Purdue University researchers have introduced a new, physically practical method called DetStorm that reveals a critical vulnerability in how autonomous vehicles process visual information.
DetStorm uses projected light patterns to flood a vehicle’s cameras with convincing, but fake, objects. This overload causes significant delays in how the system interprets its surroundings, slowing decision-making by up to 8.1 seconds and increasing the number of detected objects by 500% in some cases. The findings highlight the potential real-world risks of perception-based attacks on autonomous systems.
“Imagine if a car sees so many things at once that it freezes before deciding what to do. DetStorm pushes systems to that breaking point,” says Muller. “That delay could mean the difference between a safe stop and a crash.”
Unlike prior methods that relied on cumbersome physical patches, DetStorm uses dynamic light projection, which adapts in real time to changing environments. The attack employs a pre-generated “perturbation dictionary” and an intelligent zone-based approach to make fake objects appear where they will be most effective.
DetStorm creates a large number of false objects (colored boxes) that overwhelm an autonomous system’s perception.
The research demonstrates not only a new class of attack but also highlights the need for defense systems that don't rely solely on perception data. “Our work shows that if an attacker can delay perception, they can also delay defenses,” Muller says.
The research was led by Raymond Muller, a research assistant and now Ph.D. graduate from Purdue’s Department of Computer Science, under the guidance of Assistant Professor Z. Berkay Celik, co-director of the Purdue Security (PurSec) Lab.
The real-world connected autonomous vehicle used for experiments represents a collaboration between Purdue departments.
This work resulted from a collaboration with academic researchers, including Ruoyu Song from Purdue University, Chenyi Wang and Professor Ming Li from the University of Arizona, Yuxia Zhan from New York University, and Professor Ryan Gerdes from Virginia Tech. Industry collaborators were Jean-Philippe Monteuuis and Jonathan Petit from Qualcomm, and Yanmao Man from HERE Technologies. The project also benefited from the expertise of Professor Yiheng Feng from Purdue’s Lyles School of Civil and Construction Engineering, who, alongside students Wangzhi (George) Li and Tianheng Zhu, supported real-world autonomous vehicle experiments and provided a real-world connected autonomous vehicle to the project.
The study, enabled in part by Purdue’s high-performance Gilbreth computing servers, reflects a deep network of interdisciplinary collaboration. “This discovery is a culmination of partnerships across departments and with industry leaders, carefully fostered by Dr. Celik,” said Muller.
Funding for the research came from the National Science Foundation and the Army Research Office.
The Detstorm project is part of Purdue’s ongoing leadership in autonomous systems security. For more on this research and the PurSec Lab’s work, click here.
About the Department of Computer Science at Purdue University
Founded in 1962, the Department of Computer Science was created to be an innovative base of knowledge in the emerging field of computing as the first degree-awarding program in the United States. The department continues to advance the computer science industry through research. U.S. News & World Report ranks the department No. 8 in computer engineering and No. 16 overall in undergraduate and graduate computer science. Additionally, the program is ranked No. 6 in cybersecurity, No. 8 in software engineering, No. 13 in systems, No. 15 in programming languages and data analytics, and No. 18 in theory. Graduates of the program are able to solve complex and challenging problems in many fields. Our consistent success in an ever-changing landscape is reflected in the record undergraduate enrollment, increased faculty hiring, innovative research projects, and the creation of new academic programs. The increasing centrality of computer science in society, academic disciplines and new research activities — centered around foundations and applications of artificial intelligence and machine learning, such as natural language processing, human computer interaction, vision, and robotics, as well as systems and security — are the future focus of the department. Learn more at cs.purdue.edu.