Apple, Qualcomm, and AMD GPUs Susceptible To Putting Artificial Intelligence Data at Risk
Millions of Apple, Qualcomm, and AMD chips are susceptible to the security flaw.
- A new security flaw in popular GPUs has been found by security researchers that potentially impacts AI large language models (LLMs).
- While the new attack method requires physical access to the GPU, it opens up major implications with the growing adoption of AI tools.
Researchers from cybersecurity firm Trail of Bits have found a major security flaw in popular graphic processing units (GPUs) that enables threat actors to extract data from graphic card memory regardless of its creation method. The threat could result in significant data leakage, which is a major issue, especially for users of machine learning and large language models (LLMs).
This is particularly concerning owing to the boom in AI applications in recent months. The vulnerability allows malicious actors to eavesdrop on interactive user sessions, with a 5 to 180 megabytes exposure rate.
See More: U.S. Federal Agencies Send Out Warnings About Androxgh0st Malware Botnet
The issue known as LeftoverLocals particularly impacts GPUs from popular brands such as Imagination, Apple, AMD, and Qualcomm. Apple, Qualcomm, and Google have made patches for some devices, but others might remain affected. NVIDIA and ARM have stated that the problem has not impacted their GPUs.
The vulnerability requires a threat actor to have access to the GPU physically, but it circumvents some of the more common security measures. The issue highlights the need for AI and ML security experts to review development stacks rigorously.
What do you think about the potential threats of AI data leaks? We’d love to hear from you! Let us know your thoughts on LinkedIn, X, or Facebook. We’d love to hear from you!
Image source: Shutterstock
LATEST NEWS STORIES
