Researchers from MIT have introduced TONTOU, a new side-channel attack that exploits timer interrupts to reopen the branch predictor poisoning window. This vulnerability allows the attack to slip through existing Spectre mitigations on both Intel and AMD processors. The team demonstrated the concept with a working exploit targeting Zen 2 architecture, confirming that current defenses may be insufficient against this specific timing-based approach.
- Timer interrupts can reset or expose branch predictor states, creating a new attack vector.
- Existing Spectre mitigations may fail against TONTOU on Intel and AMD CPUs.
- Zen 2 architecture is vulnerable, as proven by the researchers' working exploit.
- Database and cloud workloads relying on shared cores are at risk of information leakage.
- Monitor for new microcode updates addressing timer-interrupt-based branch prediction flaws.
