On 2017-07-10 00:04, Kurt Seifried wrote:
>
https://www.aimlab.org/haochen/papers/npc16-overflow.pdf
>
> I really think CVE needs to consider more/better hardware coverage.
I only skimmed the first few pages, but got the impression that paper says "GPU architecture is different that modern CPUs, especially around memory layout/protection" and it isn't immediately clear to me where responsibility for any vulnerability lies. Is
it simply not possible to write memory-corruption-proof code for GPUs?
Regardless, I have no problem with issuing CVE IDs for hardware vulnerabilities. I'm just claiming that they are rare, and usually involve low-level software.
- Art