A new attack slips past the latest defenses built into your computer's processor
Daniel Truchillo and Mengjia Yan from MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) discovered a new attack called TONTOU that can bypass built-in defenses of processors. The study indicated that existing protection mechanisms become ineffective due to the time gap between the prediction of machine behavior and actual execution. Using “interrupt injection” technology, the team precisely controlled when the processor would pause, successfully causing misjudgments on four types of AMD and Intel processors and stealing memory secrets. Experiments showed that the attack could read the `/etc/shadow` file containing the root password hash in about nine minutes. The researchers have notified AMD, Intel, and Linux kernel maintainers to coordinate a response, but a secondary cleanup solution for Intel chips might be counterproductive, and blocking interrupts could lead to significant performance losses.