It would have to exist first, and then be adopted. That may never happen since CFI will be even further ahead upon release.
1
What if Microsoft adopts fine-grained CFI before this ships? They'll have no reason to adopt the CFI portion of this.
1
The return address protection will definitely be useful, although if MPK ships first it can implemented in the meantime.
1
So CET would end up being a performance optimization. MPK-based shadow stack would be similar in cost to stack canaries.
2
Ive seen numerous mitigations not ship due to single digit perf overhead. I reckon CET will drive widespread adoption
1
Maybe I'm missing something, but doesn't Microsoft currently ship software-only CET and call it CFG?
1
yes, although perf hit may not allow to make it as fine grained as wanted. Cpu support will likely help there
1
A coarse CPU implementation won't really help with that and can do type-based CFI with minimal overhead today.
1
Jun 10, 2016 · 11:47 PM UTC
2
1