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

An unprivileged Linux program can time a hardware interrupt to land in the gap between a processor sanitizing its branch predictor and the kernel using it, re-poisoning the predictor after the defense has run. MIT CSAIL researchers Daniël Trujillo and Mengjia Yan named the technique INTERRUPT INJECTION. On an AMD Zen 2 machine running Linux 6.14 with every default Spectre v2 mitigation on,

The article reports a CPU side-channel attack called INTERRUPT INJECTION that can bypass Spectre v2 mitigations on Intel and AMD systems by timing an interrupt between predictor sanitization and kernel use, enabling unprivileged local code execution to leak kernel memory on affected Linux systems.[1][2][4] On an AMD Zen 2 test system, researchers reportedly leaked arbitrary kernel memory and could read sensitive data such as /etc/shadow, showing the issue can expose secrets even with default mitigations enabled.[2][7] RealGround analysis: this is primarily a platform hardening and exposure-management issue rather than an AI-specific risk, so the closest fit is a general security readiness/advisory service for systems that may host AI workloads or sensitive data.