Software developed by assistant professor in computer science and software engineering being used in breakthrough technique to identify leukemia cells
Computer Science and Software Engineering
By Joe McAdory
As millions of Americans rely on AI-powered security cameras to protect their homes, new research from Auburn Engineering reveals a critical vulnerability: inexpensive laser pointers can disable the devices and bypass their security features.
Led by Yazhou Tu, assistant professor in the Department of Computer Science and Software Engineering, the research team found that smart doorbells can be defeated by low-cost laser attacks.
“Cybersecurity nowadays extends farther than computers, websites or servers,” said graduate student Eftakhar Ahmed Arnob. “We must also consider cyber-physical systems, including smart homes and autonomous technologies.”
The vulnerability stems from a design flaw in the cameras’ optical systems. When a laser is directed at the lens, light reflects internally, creating distortions that confuse AI detection systems.
As a result, the cameras fail to recognize motion or people — and do not alert homeowners that their view is being blocked.
Tu’s team is developing defenses based on adversarial optical physics rather than traditional deep learning models. The approach identifies abnormal light patterns caused by laser interference, eliminating the need to train systems on every possible attack scenario.
“We don’t have a silver bullet to make real-world systems absolutely secure,” Tu said.
Researchers are continuing to collect data and working with manufacturers to implement both short- and long-term solutions, including software updates, algorithm improvements and potential hardware changes.