Google awards $250K bounty for Linux KVM guest escape vulnerability discovered after 16 years

A previously unknown vulnerability in the Linux Kernel-based Virtual Machine (KVM) allows untrusted guest virtual machines to break out of isolation and execute code with root privileges on the host, Google’s bug-bounty program confirmed. The flaw, tracked as CVE-2026-53359, resides in KVM’s shadow MMU emulation and has gone undetected for 16 years, according to researcher Hyunwoo Kim, who named it Januscape.

Kim described Januscape as a use-after-free vulnerability that corrupts the host kernel’s shadow page table by triggering guest-side actions alone. An attacker with root access inside a rented cloud VM could exploit the flaw to crash the host kernel (denial of service) or achieve remote code execution on the host and all co-located guest VMs. Kim released a proof-of-concept that crashes the host OS but withheld a full escape exploit, stating it would not be released “until the very distant future.”

The vulnerability affects KVM on both AMD and Intel processors and can be exploited even in cloud environments that deploy custom virtualization stacks, because the flaw is not in QEMU’s memory translation path. Exploitation requires the guest VM user to already have root privileges within the guest.

Google awarded a $250,000 bounty for reporting CVE-2026-53359 through its kernelCTF program. A separate Linux vulnerability, CVE-2026-43499 (GhostLock), also received a patch this week. Discovered by researchers at Nebula Security using the company’s AI-assisted scanner Vega, GhostLock is a use-after-free flaw in the kernel’s futex priority-inheritance machinery that had existed for 15 years.

GhostLock enables users with limited rights to escalate privileges to root by exploiting a cleanup error in the futex subsystem, ultimately tricking the kernel into executing attacker-controlled code as root. The flaw carries a severity rating of 7.8 out of 10 and earned the researchers a $92,337 bounty from Google’s kernelCTF program.

Both vulnerabilities have been patched in the mainline Linux kernel. Linux distributions are expected to propagate the fixes to end users, who should verify their systems have received the updates.