University of California, Riverside scientists have determined how hackers can exploit graphics processing units (GPUs) to expose private data.
The team reverse-engineered an Nvidia GPU to present three general side attacks on both graphics and computational stacks, in which victims must obtain malicious spyware embedded in a downloaded app.
The first strategy has the victim open the malware, with OpenGL generating an agent that deduces browser behavior from GPU memory utilization and infers all allocation events to track online behavior.
The second hack targets passwords, because as the user types a character, the password textbox is uploaded to GPU as a texture to be rendered; monitoring the interval time of consecutive memory allocation events reveals the number of password characters and inter-keystroke timing.
The third attack is cloud-focused, with a malicious computational workload launched on the GPU to operate alongside the victim's app. Machine learning techniques infer the target's secret neural network structure.
From University of California, Riverside
View Full Article
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found