Loading vulnerability…
Zapscape is a guest-to-host escape vulnerability in the Linux kernel’s KVM/x86 virtualization subsystem, affecting the shadow MMU fault-handling path. The flaw stems from checking whether the current shadow MMU root is invalid or obsolete before calling the logic that makes MMU pages available. During that reclamation step, KVM can zap an in-use root and mark it invalid, but the vulnerable path may continue and map memory through that now-invalid root. Because newly created child shadow pages inherit properties from the parent, KVM can create child shadow pages beneath an invalid root and place invalid pages onto the active MMU page list, violating internal invariants. This can lead to list corruption, dangling references, and a use-after-free condition in the recursive shadow-page zap path. The vulnerability affects KVM/x86 on both Intel and AMD systems and has been publicly demonstrated as a full guest-to-host escape resulting in host root code execution.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos (5 hidden).
This repository is a research-oriented proof-of-concept for CVE-2026-64561, named Zapscape, a guest-to-host escape in Linux KVM/x86 caused by a use-after-free in the shadow MMU recursive zap path during MMU page quota reclaim. The repository is small and mostly documentation: a top-level README, a nested Zapscape-V4bel directory with a detailed README, a long technical write-up, SVG assets, and one executable helper script (qemu.sh). No actual exploit source code such as poc.c is included in the provided contents; instead, the repository documents the exploit chain and how to boot a vulnerable lab environment. The exploit capability described is severe: an attacker controlling a nested guest can corrupt host KVM shadow MMU state and escalate to host kernel code execution. The demonstrated end result is creation of /Zapscape on the host as a root-owned file, proving guest-to-host escape. The write-up further states the chain obtains KASLR slide information and abuses kernel paths including log_wait, SRCU workqueue, and usermode helper mechanisms to reach host-side privileged execution. Targeting is Linux KVM/x86 in the affected range from commit f95eec9bed76 through the fix at 2abd5287f083, with the public demo focused on AMD nested SVM/NPT on Linux 7.1.3. Intel is also affected under narrower conditions requiring both EPT page walk lengths 4 and 5 exposed to L1. The exploit is relevant both as a hypervisor escape in multi-tenant cloud environments and as a local privilege escalation path on hosts where /dev/kvm is accessible to unprivileged users. Operationally, the included qemu.sh script is the main entry point. It launches qemu-system-x86_64 with TCG, EPYC CPU emulation, nested AMD virtualization flags (+svm,+npt), 2 GB RAM, 2 vCPUs, no networking, and kernel parameters enabling kvm_amd.nested=1 and kvm_amd.npt=1. This script is a lab harness rather than the exploit itself. Overall, the repository is a credible exploit PoC/documentation package, not a detection script and not obviously weaponized; it is best classified as POC maturity.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
98 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
A use-after-free virtualization escape vulnerability in Linux KVM's x86 shadow MMU that can allow an attacker with kernel-level privileges in a nested L1 guest to execute code on the host under specific conditions.
A Linux kernel KVM/x86 guest-to-host escape vulnerability in KVM's shadow MMU caused by a use-after-free in the recursive zap path, enabling a malicious guest to potentially gain root control of the host.
A guest-to-host escape vulnerability in the Linux kernel KVM x86 shadow MMU logic that can let an attacker controlling a guest VM corrupt the host kernel and achieve host kernel code execution, especially in nested virtualization scenarios.
A use-after-free vulnerability in the Linux kernel KVM virtualization subsystem that can allow local privilege escalation to root, likely by an attacker with access to /dev/kvm. It affects Intel and AMD systems using shadow page tables/MMU virtualization paths.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.