Nebula Security publishes GhostLock exploit for 15-year Linux kernel flaw

Nebula Security published exploit code for GhostLock, a 15-year-old Linux kernel vulnerability (CVE-2026-43499) that allows local privilege escalation and container escape. The flaw, introduced in Linux 2.6.39 and fixed in 7.1, was discovered by Nebula's AI-assisted Vega system, which earned a $92,337 Google bounty. The exploit highlights the growing role of AI in finding long-hidden kernel bugs.

Nebula Security @nebusecurity https://x.com/nebusecurity published technical details and exploit code for GhostLock, a Linux kernel vulnerability tracked as CVE-2026-43499, in a July 8th thread on X https://x.com/nebusecurity/status/2074663573742338256 and a July 7th research writeup https://nebusec.ai/research/ionstack-part-2/ . https://x.com/nebusecurity/status/2074663573742338256 https://x.com/nebusecurity/status/2074663573742338256 The disclosure is current; the bug itself is older. Nebula says GhostLock was introduced in Linux 2.6.39 and fixed in Linux 7.1 after sitting in the kernel for roughly 15 years. The National Vulnerability Database lists the issue in kernel/locking/rtmutex.c , describes it as a use-after-free flaw in the rtmutex path, and records kernel.org as the source. NVD gives the kernel.org CVSS 3.1 score as 7.8, high, with local attack vector, low attack complexity, low privileges required, and no user interaction. Nebula says its Vega system found GhostLock and that the lab turned it into a 97% stable local privilege-escalation and container-escape exploit. Google rewarded Nebula $92,337 for the submission through kernelCTF, according to Nebula's writeup. That reward figure has not been independently posted by Google in the public sources we found, so it should be read as Nebula's account of the bounty. The vulnerability is a reminder of what security startups are now selling around AI-assisted research: proof that an automated or semi-automated pipeline can find kernel-class bugs that were missed for years. Nebula describes Vega https://nebusec.ai/vega/ as an AI security pipeline for codebases and claims on its product page that it has produced 730 validated findings, 698 Linux kernel bugs, 8 Chrome zero-days, and 90 public CVEs. Those are Nebula's own metrics, but GhostLock gives the pitch a concrete artifact: a CVE, a kernel patch trail, a public exploit repo, and a claimed Google bounty. The root cause is narrow. NVD says remove waiter was used both in normal slowlock paths and in proxy-lock rollback through rt mutex start proxy lock when invoked from futex requeue . In the proxy path, the waiter task is not the current task, but the code operated on current . That mismatch leaves the waiter task's pi blocked on state uncleared, creating a dangling pointer that can be used around a use-after-free condition. Nebula's explanation turns that kernel bookkeeping error into an attack chain. GhostLock gives an unprivileged local attacker a dangling kernel pointer to stack memory through regular threading syscalls, then a pointer write primitive, then a path to hijack a function table and reach root, according to the writeup. Nebula says the affected range runs from v2.6.39-rc1 to v7.1-rc1 , with CONFIG FUTEX PI=y as the only requirement and no capabilities or user namespaces needed. The most important constraint is that GhostLock is local. It is not a drive-by remote compromise on its own. Its value comes when an attacker already has code execution inside a Linux environment, such as a container, a browser sandbox, a mobile process, or a lower-privileged account, and needs the kernel bug to cross the next boundary. That is why Nebula framed GhostLock as part of IonStack, its broader browser-to-kernel chain. Nebula's IonStack demo site https://rootme.nebusec.ai/ describes IonStack as a browser-to-kernel full-chain remote code execution path on Android 17 and points users to the same CyberMeowfia open-source code base. Nebula also wrote in the GhostLock thread that it plans to publish how it uses GhostLock to root Android and bypass KASLR and KCFI. That Android writeup has not been published in the source material provided for this story. The patch state is mixed across downstream distributions. NVD records the CVE as published on May 21st, 2026, and lists kernel.org patch references. Debian's security tracker lists Debian bookworm's linux source package fixed in 6.1.176-1, trixie fixed in 6.12.86-1, and sid fixed in 7.1.3-1, while some bullseye and earlier bookworm entries remain marked vulnerable. Ubuntu's tracker, last updated July 6th, lists CVE-2026-43499 as high priority with a 7.8 CVSS 3 score and shows the main linux package fixed for Ubuntu 26.04 LTS in 7.0.0-27.27 while several older supported releases remain marked vulnerable or work in progress. That status table matters because Nebula's broad claim that "everything around you" running Linux is affected is technically bounded by kernel lineage and patch adoption. The practical question for operators is whether a given machine has received the rtmutex fix through its vendor's kernel package. A patched system is not in the same position as an older Ubuntu, Debian, Android, IoT, or custom-kernel deployment that still carries the vulnerable code. Nebula also published the GhostLock exploit and PoC https://github.com/NebuSec/CyberMeowfia/tree/main/IonStack/CVE-2026-43499 under its CyberMeowfia repository. GitHub showed the repository as public, with folders for exploit and poc , and 83 stars when reviewed. Publishing working exploit code increases the urgency for lagging patch channels, even when the flaw requires local execution. The timing tells a clean disclosure story. Nebula says it reported the bug and sent a draft patch to security@kernel.org on April 18th, the bug was fixed on April 20th, the first backport landed on May 4th, Google acknowledged the kernelCTF submission and paid the claimed bounty on June 30th, and Nebula published the writeup on July 7th. Nebula's about page https://nebusec.ai/about/ says Nebula follows responsible disclosure by reporting privately, giving vendors time to patch, then publishing once a fix has shipped. The business read is straightforward. Nebula is using an old kernel bug with a measurable bounty and public exploit chain to sell confidence in Vega. That is a stronger proof point than a generic AI security claim because kernel exploitation leaves less room for marketing language: the code either escapes, earns a CVE, and survives maintainer scrutiny, or it does not. GhostLock has moved through the kernel CVE process and into downstream vendor trackers. The remaining risk sits with slow-moving fleets that treat kernel patching as an operations problem until exploit code turns it into a security deadline.