Loading story…
CVE-2026-74580, dubbed Vghost, is a Linux vhost vulnerability that lets local users with access to vhost device nodes read and write arbitrary kernel memory, bypass KASLR, and escalate to root. Reportedly affecting kernels from Linux 4.11 through 7.2-rc4, the flaw has a reported CVSS score of 8.8. Cached virtual-ring memory translations remain valid after ring reconfiguration, while missing range validation allows stale mappings to reach memory-copy operations. Under a specific virtual IOMMU and device-IOTLB configuration, an untrusted guest can also trigger out-of-bounds writes into the host’s QEMU process, potentially hijacking execution or crashing it; this guest-to-host path is configuration-dependent.
Tencent Zhuque Lab discovered the flaw with assistance from Corvus AI and published technical details and working demonstrations on October 12, following a reported fix release on August 3. A public x86-64 proof of concept targets the vhost-vsock device and demonstrates temporary modification of the kernel’s module-loader path. The repository identifies commit de845981da67 as the fix, and AWS has a corresponding CVE advisory entry. No public exploitation has been reported. Administrators should prioritize patched kernels and reboot affected systems, especially virtualization hosts. Restricting vhost device access is an interim mitigation for local attacks, but the reporting identifies no interim workaround for the guest-to-host attack.
Get the actors, campaigns, and ATT&CK mapping behind it.
9 events from the most recent confirmed update back to the earliest known activity.
Tencent published a full technical write-up with working demonstrations of local privilege escalation and guest-to-host attacks. The research showed how stale vring mappings enable arbitrary kernel-memory access locally and out-of-bounds writes into the host's QEMU process from a guest.
On August 21, 2026, Greg Kroah-Hartman announced CVE-2026-74580 on linux-cve-announce, describing stale vhost vring metadata mappings that can permit memory accesses outside IOTLB-mapped regions. The advisory identified fixes in eight kernel releases and recommended upgrading to the latest stable kernel rather than cherry-picking patches.
Tencent released a fix for Vghost. Commit de845981da67 clears cached metadata mappings whenever a virtual ring is reconfigured.
A Linux 5.9 change deferred address checking and removed a check that would otherwise have caught stale translations, contributing to the conditions behind Vghost.
A Linux 4.11 change introduced cached mappings for virtual-ring metadata and a validation shortcut. These mechanisms later contributed to the stale-translation vulnerability known as Vghost.
Red Hat, Amazon Linux, and Oracle Linux shipped fixes for Vghost, and Ubuntu fixed its 22.04 kernels. The source does not specify when these fixes were released.
The security-research repository provides an x86-64 proof of concept targeting /dev/vhost-vsock that temporarily modifies modprobe_path, verifies the change, and restores its original value. It contains no payload, does not spawn a root shell, and does not exercise the guest virtqueue-reset attack path.
Greg Kroah-Hartman committed Jun Yang's vhost metadata-cache fix to the Linux stable tree as commit 13fa6f32a56a386a82bd7451644c494beed034af, referencing upstream commit de845981da67. The patch resets cached metadata mappings during vring reconfiguration, forcing the new addresses through IOTLB validation.
Tencent Zhuque Lab discovered CVE-2026-74580 in the Linux vhost subsystem with assistance from its Corvus AI kernel-research system. The vulnerability permits local privilege escalation and, under a specific virtualization configuration, guest-to-host attacks.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
Get the adversaries, campaigns, and ATT&CK mapping behind this technique, with detections ready to deploy.
9 references tracked. Mallory keeps watching after this page renders.
thehackernews.com
Open sourcecve.org
Open sourcegithub.com
Open sourcematrix.tencent.com
Open sourceexplore.alas.aws.amazon.com
Open sourcelore.kernel.org
Open sourcegit.kernel.org
Open sourcegit.kernel.org
Open sourcegit.kernel.org
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
A newly disclosed KVM/x86 vulnerability, **CVE-2026-64561** or **Zapscape**, allows a privileged L1 guest to break nested virtualization isolation and achieve guest-to-host escape on Linux systems. The bug is a use-after-free in KVM/x86 shadow MMU emulation tied to the recursive shadow-page zap path and stale-root check ordering during guest-triggered page fault handling, enabling corruption of host kernel shadow pages and potential host code execution. Researcher **Hyunwoo Kim** reported the flaw to the Linux kernel security team and published a proof-of-concept showing creation of a root-owned file on the host. The issue affects both **Intel** and **AMD** hosts and is especially serious for environments that run untrusted guests with nested virtualization enabled, including multi-tenant public cloud infrastructure. On Intel, exploitation requires exposing both EPT page walk lengths 4 and 5 to the L1 guest, while no equivalent constraint was noted for AMD. The vulnerability affects Linux KVM versions from **5.9 onward** across the disclosed commit range, and a fix has been merged upstream into mainline Linux, with users advised to deploy patched stable kernels or vendor-backported updates; no in-the-wild exploitation has been claimed.
Aug 7, 2026
A public disclosure detailed **Zapscape**, tracked as `CVE-2026-64561`, a guest-to-host escape vulnerability in **KVM/x86** caused by a use-after-free in shadow MMU emulation. The flaw occurs in the recursive shadow-page "zap" path and can be triggered through guest-side actions alone, allowing corruption of host kernel shadow pages and breaking guest-host isolation. The issue affects both **Intel** and **AMD** hosts; on Intel, exploitation requires that EPT page walk lengths 4 and 5 are both exposed to the L1 guest, while no equivalent constraint was noted for AMD. The disclosure warns that systems running untrusted guests with nested virtualization enabled are at particular risk, including multi-tenant x86 public cloud environments. Hyunwoo Kim reported the bug to the Linux kernel security team, and the issue has been patched in mainline Linux. The affected range spans from commit `f95eec9bed76` through `2abd5287f083`, covering code introduced in 2020 and remaining exposed until the July 2026 fix referenced on kernel.org.
Aug 7, 2026
**CVE-2026-89775** is a guest-to-host escape vulnerability in Linux KVM on ARM64 systems with nested virtualization enabled. A stage-1 page-table walk-level truncation can cause a range-size calculation to return zero; VNCR pseudo-TLB invalidation then skips invalidation, potentially leaving a freed host page mapped and writable at a fixed host-kernel address. A malicious guest can obtain 64-bit read/write access to host kernel memory and may be able to execute code on the host. The issue affects multi-tenant ARM64 cloud deployments that expose nested-virtualization-capable instances and local systems where unprivileged users can access `/dev/kvm`. Nested virtualization is experimental, disabled by default, and requires Armv8.4 `FEAT_NV2` hardware, limiting exposure. Upstream fixed the flaw in commit `8053393680d4`; fixes are available in Linux 6.18.51, 7.2.5, and 7.3-rc1. No public exploit or active exploitation had been reported.
Sep 23, 2026