Scientists at the University of California have reported for the first time about possible attacks from hackers on the graphics processing unit (GPU). They can spy steal passwords, spy on web activities, and access into cloud-based applications. The Marlan and Rosemary Bourns College of Engineering computer science doctoral student Hoda Naghibijouybari and post-doctoral researcher Ajaya Neupane, along with Associate Professor Zhiyun Qian and Professor Nael Abu-Ghazaleh, reverse engineered Nvidia GPU to demonstrate 3 attacks on both graphics and computational stacks, and across them.
The group believes these are the first reported general side-channel attacks on GPUs. All 3 attacks need the victim to first acquire a malicious program embedded in a downloaded app. The program is designed to spy on the victim’s computer. Web browsers use GPUs to render graphics on desktops, laptops, and smartphones. GPUs are also used to accelerate applications on the cloud and data centers. Web graphics can disclose user information and activity. Computational workloads boosted by the GPU include applications with important data or algorithms that could be exposed by the new attacks.
The first attack monitors user activity on the web. The victim opens the malicious app, as it uses OpenGL to create a spy to infer the behavior of the browser. Every website has a unique track in terms of GPU memory utilization because the different number of objects and various object sizes to be rendered. The signal is consistent across loading the same website many times. The second attack follows the author’s extracted user passwords. Each time the user types a character, the whole password textbox is uploaded to GPU as a texture to be rendered. The third attack targets a computational application in the cloud. The attacker launches a malicious computational workload on the GPU. It operates alongside the application of the victim.