MIT researchers demonstrate interrupt-based bypass of processor defenses

MIT CSAIL researchers reported a class of speculative-execution attacks called TONTOU that uses precisely timed interrupts to contaminate processor prediction machinery after it has been cleared. They tested four Intel and AMD processor generations and demonstrated a working exploit on an AMD system running a current Linux kernel. The exploit read protected memory at about five bytes per second and copied the root-password hash file in half of attempts. AMD released a mitigating patch.