Branch Target Reuse Spectre-v2 Attack Leaks Linux Kernel Memory
Vulnerability Disclosure · 15 sources · Updated 2mo ago · First seen Sep 29, 2026

Researchers at VUSec and Scuola Superiore Sant’Anna disclosed Branch Target Reuse (BTR), a Spectre-v2-class attack that abuses stale indirect-branch predictor entries after JIT-generated code is freed and its cache region reused. The resulting speculative execute-after-free condition can redirect transient execution and expose secrets through cache side channels. Tests found the issue across evaluated Intel, AMD, and Arm CPUs and in JIT environments including Linux cBPF, Firefox SpiderMonkey, and Oracle GraalVM.
Two end-to-end exploits against Linux’s cBPF JIT leaked arbitrary kernel memory on fully patched Intel systems with default mitigations enabled, recovering a root password hash within minutes at roughly 8 bytes per second. Linux has released and backported cBPF hardening tracked as CVE-2026-64507 and CVE-2026-64508; mitigations use indirect branch prediction barriers on x86. GraalVM is randomizing JIT code-cache locations, while Mozilla is prioritizing Firefox site isolation after evaluating IBPB-based defenses. Researchers warned that existing hardware lacks a general mechanism to synchronize branch-prediction state with modified runtime code.


