INTERRUPT INJEC Bypasses Spectre v2 on Intel, AMD
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- MIT CSAIL researchers Daniël Trujillo and Mengjia Yan discovered a new attack technique, named INTERRUPT INJEC, that bypasses Spectre v2 defenses on both Intel and AMD CPUs.
- The technique exploits the window between a processor sanitizing its branch predictor and the kernel using it, allowing an attacker to re-poison the predictor after the defense has already run.
- An unprivileged Linux program is sufficient to carry out the attack — no elevated privileges or kernel access are required, according to the source.
- The attack works by timing a hardware interrupt to land precisely in that sanitization gap, re-corrupting the predictor state before the kernel consumes it.
- Both Intel and AMD processors are affected, indicating the vulnerability is architectural rather than vendor-specific.
Why it matters: Spectre v2 mitigations are core to securing branch prediction across nearly every modern Intel and AMD CPU shipped in recent years; if an unprivileged userland program can sidestep them by timing an interrupt, cloud tenants, shared hosts, and browsers handling cross-origin code gain a realistic new side-channel primitive that bypasses defenses vendors spent years deploying.
