Linux KVM guest-host escape flaw CVE-2026-99999 rated critical
Researcher Paulos Yibelo names the critical Linux KVM vulnerability PwnEscape, warning it allows guests to break isolation via malformed IPv4 packets.
Key points
- The flaw CVE-2026-99999 receives a CVSS score of 9.3, classifying it as critical.
- Attackers can trigger the guest-host escape by sending malformed IPv4 packets.
- The issue stems from how the kernel handles checksum metadata in the IPv4 stack.
Paulos Yibelo identified a critical guest-host escape vulnerability in the Linux KVM hypervisor, assigning it CVE-2026-99999 and the name PwnEscape. According to The Register, Yibelo made the claim in early October, warning that the defect represents a new attack vector for Linux systems used in cloud environments.
The vulnerability carries a CVSS score of 9.3, marking it as a critical severity issue. Yibelo’s research demonstrates that an attacker inside a virtual machine can break out to the host system by transmitting a few malformed IPv4 packets. This breach violates the core isolation principle of virtualization, which is designed to keep guest environments separate from the host and other guests.
Technical details of the escape
The flaw originates in how the Linux kernel processes network packet metadata. As explained by Red Hat engineer Michael S. Tsirkin in the Linux kernel stable tree, packets with bad checksum metadata can corrupt the network header. Tsirkin noted that the function `skb_checksum_help` can cause significant mischief when such corrupted data enters the IPv4 stack.
SOC Prime provided analysis detailing the specific mechanics of the exploit. The firm stated that Linux allows packets to carry metadata indicating where checksum operations begin and where results are written. This optimization is standard for checksum offloading but becomes dangerous when guest-controlled metadata influences writes after earlier validation stages have passed.
Impact on cloud infrastructure
Public and private clouds rely heavily on the isolation provided by hypervisors like KVM. If guest-host isolation is compromised, an attacker could potentially access all virtual machines on a single host or within a pool of virtualized hardware. The Register reported that such an attack could grant access to thousands of VMs.
The risk extends to containerized workloads as well. Many developers run containers inside micro-VMs to maintain isolation. A successful PwnEscape attack on the host could therefore compromise these containers, undermining the security benefits of using micro-VMs for workload separation.
What should you do?
Security professionals should monitor for vendor patches addressing CVE-2026-99999. Until a fix is confirmed, organizations using Linux KVM should review their network segmentation policies. Isolating critical workloads on separate physical hosts may reduce the blast radius of a potential escape.
What to do and how to stay safe: Linux
- Monitor official Linux kernel and hypervisor vendor channels for patches addressing CVE-2026-99999.
- Review network segmentation to ensure that compromised guests cannot easily reach other critical assets.
- Audit virtual machine configurations to limit unnecessary network access for guest systems.
- Verify that checksum offloading features are configured securely on host network interfaces.
Step-by-step guide: Risk-Based Vulnerability Management: Prioritize Fixes by Actual Danger
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What is the CVSS score for the PwnEscape vulnerability?
The vulnerability CVE-2026-99999 has a CVSS score of 9.3, which is rated as critical.
How do attackers trigger the guest-host escape?
Attackers can trigger the escape by sending malformed IPv4 packets to the host system.
Which component of the Linux kernel is affected?
The flaw affects the Linux KVM hypervisor and the IPv4 stack, specifically involving checksum metadata handling.
