Just look at the amount of brilliant, hard work in the abstract. That is the reason I don’t trust tactical mitigations. After smashing others, they say they’ll propose some mitigations. I didn’t read them. I didn’t because the next paper will be similarly clever people with a lot of time on their hands smashing their mitigations. It goes on and on. I still skim them in case something breaks my recommendations or others in high-assurance security. Do they have a section on that?
“Stopping PIROP primitives. A practical way to stop PIROP attacks is to remove the necessary primitives. For
example, to eliminate relative memory write primitives, we may rely on memory safety solutions [59]–[64] or, alternatively, data-flow integrity solutions [65], [66]. However, these solutions incur nontrivial performance overhead, reducing the performance benefits of ASLR.”
(wipes sweat off forehead) Oh, it doesn’t work on memory safety and data-flow integrity. Those are what I was just recommending on a text editor thread. Glad the fundamental techniques are still standing. The implementations still need more review, though, just in case.
Just look at the amount of brilliant, hard work in the abstract. That is the reason I don’t trust tactical mitigations. After smashing others, they say they’ll propose some mitigations. I didn’t read them. I didn’t because the next paper will be similarly clever people with a lot of time on their hands smashing their mitigations. It goes on and on. I still skim them in case something breaks my recommendations or others in high-assurance security. Do they have a section on that?
“Stopping PIROP primitives. A practical way to stop PIROP attacks is to remove the necessary primitives. For example, to eliminate relative memory write primitives, we may rely on memory safety solutions [59]–[64] or, alternatively, data-flow integrity solutions [65], [66]. However, these solutions incur nontrivial performance overhead, reducing the performance benefits of ASLR.”
(wipes sweat off forehead) Oh, it doesn’t work on memory safety and data-flow integrity. Those are what I was just recommending on a text editor thread. Glad the fundamental techniques are still standing. The implementations still need more review, though, just in case.