Researchers Use New CPU Attack to Steal Linux Root Password Hash From Memory
Researchers' Branch Target Reuse attack leaks Linux root password hashes via stale CPU predictions.
VUSec researchers disclosed Branch Target Reuse, a Spectre-v2 variant that abuses stale branch-target buffer entries after JIT code is freed and its memory is reused. Proof-of-concept exploits against the Linux kernel cBPF JIT leaked arbitrary kernel memory at about 8 bytes per second and recovered a root password hash from an su process in roughly three minutes on Raptor Cove and five minutes on Lion Cove Intel CPUs. The local attack worked despite cBPF constant blinding and is not a remote network exploit. Linux fixes tracked as CVE-2026-64507 and CVE-2026-64508 insert an Indirect Branch Prediction Barrier when BPF JIT memory is reused.
- Branch Target Reuse abuses stale branch-target buffer entries after JIT code is freed.
- Linux cBPF proofs leaked kernel memory at about 8 bytes per second.
- Root password hashes were recovered in about three to five minutes on Intel CPUs.
- CVE-2026-64507 and CVE-2026-64508 add an IBPB during BPF JIT memory reuse.
- The attack requires local code and is not a remote network exploit.
- Organizations
- VUSec
Vulnerabilities mentioned
All| CVE | Vulnerability | CVSS | EPSS | Signals | Published |
|---|---|---|---|---|---|
CVE-2026-64507+1 related CVE | Linux kernel: x86/bugs: Enable IBPB flush on BPF JIT allocation Enable hardening against JIT spraying when Spectre-v2… | 5.5 | <1% | PoC |
From the original article
3 min read at cybersecuritynews.com
Branch Target Reuse (BTR) is a newly disclosed variant of the Spectre-v2 attack that leverages stale CPU branch-prediction data to leak sensitive information from Linux memory, including root password hashes. This attack specifically targets just-in-time (JIT) compilers found in the Linux kernel, web browsers, and various language runtimes. BTR takes advantage of a vulnerability that […]
By Abinaya · cybersecuritynews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.