So, I saw that you guys were working on getting GPU acceleration working in the VMs. I had an idea on potentially how to make this more secure and make GPU-based VM escapes and other GPU based attacks more difficult.
The solution you'd like
You could run the GPU API calls through a GPU call sandbox such as the ANGLE library (you might have to tweak it a bit), which validates them and ensures that GPU API calls will not result in undefined or malicious behavior.
The value to a user, and who that user might be
This would provide an additional protection for virtual machines that you want to use GPU acceleration for, but may not trust 100%.
Completion criteria checklist
(This section is for developer use only. Please do not modify it.)
The problem you're addressing (if any)
So, I saw that you guys were working on getting GPU acceleration working in the VMs. I had an idea on potentially how to make this more secure and make GPU-based VM escapes and other GPU based attacks more difficult.
The solution you'd like
You could run the GPU API calls through a GPU call sandbox such as the ANGLE library (you might have to tweak it a bit), which validates them and ensures that GPU API calls will not result in undefined or malicious behavior.
The value to a user, and who that user might be
This would provide an additional protection for virtual machines that you want to use GPU acceleration for, but may not trust 100%.
Completion criteria checklist
(This section is for developer use only. Please do not modify it.)