Bottom line: MIT researchers demonstrate with INTERRUPT INJECTION that an unprivileged Linux process can bypass existing Spectre v2 protections on Intel and AMD CPUs by precisely timing a hardware interrupt.
Researchers at MIT CSAIL have presented a technique called INTERRUPT INJECTION that allows an unprivileged Linux program to bypass existing Spectre v2 defense mechanisms. The attack exploits a time window between the processor’s cleansing of the branch predictor and its subsequent use by the kernel.
Researchers Daniël Trujillo and Mengjia Yan have developed the attack technique INTERRUPT INJECTION. It shows that an unprivileged process under Linux can time a hardware interrupt so precisely that it falls into the short gap between the CPU’s sanitization of the branch prediction state and its subsequent use by the kernel. Within this window, it becomes possible to re-poison the branch predictor even though the actual protective measure has already been executed. The attack was demonstrated on an AMD Zen 2 machine running Linux 6.14, with all Spectre v2 mitigations enabled by default active.
What is particularly relevant for CISOs is that the vulnerability does not lie in a faulty implementation of individual mitigations, but rather in the temporal interplay between hardware interrupt handling and the existing Spectre v2 countermeasures themselves. Accordingly, systems with Intel and AMD processors on which the usual Spectre v2 protection mechanisms were considered sufficient are affected. Since Spectre v2 attacks can generally be used to read memory contents across process and privilege boundaries, this potentially affects multi-tenant environments, virtualized infrastructures, and multi-tenant cloud systems in which multiple workloads share the same physical CPU.
The original source does not indicate whether CVE identifiers have already been assigned, which vendor patches are in preparation, or which specific processor generations beyond AMD Zen 2 are affected. Security officers should monitor the publication of Trujillo and Yan’s research paper as well as any advisories from Intel and AMD in order to obtain details on attack prerequisites, affected model series, and available firmware or kernel patches before specific countermeasures can be prioritized.
Source: thehackernews.com · Published August 6, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.8.3.