News › security
By Zayden R., August 7, 2026
The newly disclosed Zapscape vulnerability (CVE-2026-64561) allows guest-to-host escapes in KVM/x86 environments, threatening cloud security. Engineers should be aware of potential risks, especially in multi-tenant x86 public clouds.
A significant security vulnerability, dubbed Zapscape (CVE-2026-64561), has been disclosed, affecting KVM/x86 environments. Discovered by Hyunwoo Kim, this flaw allows a guest to escape to the host and execute commands with root privileges. This is particularly concerning for multi-tenant x86 public clouds where guest-host isolation is critical.
Zapscape is a use-after-free vulnerability found in the shadow MMU emulation of KVM/x86. It specifically targets the recursive zap path that occurs when shadow pages are reclaimed. By exploiting this flaw, an attacker can corrupt the host kernel's shadow page through guest-side actions, bypassing the isolation intended to protect the host from guest operations.
The proof of concept (PoC) for this exploit is designed to target AMD processors and is best tested under QEMU TCG for safety. The PoC involves a setup with Linux 7.1.3 and KVM_AMD on an x86_64 CPU, emulated by QEMU TCG. The process, which manipulates shadow pages, demonstrates the vulnerability without being a direct threat in a cloud environment. However, to weaponize it, actions performed by the PoC need to be integrated into a guest kernel module.
While this PoC is not immediately usable in a real-world cloud environment, it highlights a critical security gap that could be exploited if not addressed. Engineers managing KVM/x86 hosts should assess their systems, particularly if they accept untrusted guests or use nested virtualization. Patching and monitoring for updates will be crucial to mitigate the risks posed by Zapscape.
The Linux Camp teaches these topics as hands-on labs on real virtual machines, verified as you type.