New Interrupt Injection Attack Can Bypass Spectre v2 Defenses on Intel and AMD CPUs

Researchers have discovered a new attack method called Interrupt Injection that can bypass Spectre v2 defenses on Intel and AMD CPUs. This technique involves timing a hardware interrupt to occur between the processor sanitizing its branch predictor and the kernel using it, allowing the predictor to be re-poisoned after defenses have been applied. The attack was demonstrated on an AMD Zen 2 machine running Linux with all default Spectre v2 mitigations enabled. The Interrupt Injection technique was identified by MIT CSAIL researchers Daniël Trujillo and Mengjia Yan. The attack exploits a vulnerability that can be triggered by an unprivileged Linux program.