QEMU CVE-2026-48914

Exploitation requires the attacker to hold high-privilege (root or equivalent administrator) access inside the guest virtual machine - this is confirmed by the CVSS PR:H metric. The guest must be configured to use a virtio-blk device with SCSI pass-through capability enabled, as the flaw is triggered specifically via a malformed virtio-blk SCSI request; standard virtio-blk block I/O without SCSI is not confirmed as a trigger path. The attack is local (AV:L) in that it originates from within the guest, targeting the host QEMU process. Exploitation is limited by: (1) the requirement for guest root/admin privileges - unprivileged guest users cannot submit raw virtio-blk SCSI commands; (2) the need for virtio-blk with SCSI support to be provisioned for the VM; and (3) no network-facing attack surface - the attacker must already have a foothold with elevated rights inside the target guest.

The CVSS 3.1 base score of 6.7 (AV:L/AC:L/PR:H/UI:N/S:C/C:N/I:L/A:H) accurately reflects the constrained but meaningful risk profile. Attack vector is local because exploitation originates from within a guest VM; however, Scope:Changed (S:C) captures the critical guest-to-host boundary crossing. The High privilege requirement (PR:H) is the most significant limiting factor - only a guest user with administrative or root-level access inside the VM can submit raw virtio-blk SCSI commands. Availability impact is High (A:H) because a successful exploit terminates the QEMU process, which equates to an unplanned shutdown of the guest VM and disruption to host-managed workloads. Integrity impact is Low (I:L) reflecting the OOB write's ability to corrupt adjacent host heap memory, though no confidentiality impact is assessed (C:N). EPSS data was not provided in the source intelligence, limiting probabilistic exploitation forecasting. No KEV listing and no public POC are noted, which together suggest this is not yet opportunistically exploited. In cloud and virtualization hosting environments where tenants share physical hosts, this is a meaningful risk: a malicious or compromised tenant with guest root could disrupt co-hosted VMs by forcing a QEMU crash. In single-tenant or trusted-guest environments the practical risk is substantially lower.

An attacker who has obtained root or administrator-level access inside a guest virtual machine running on a RHEL or OpenShift host constructs a malformed virtio-blk SCSI request with an oversized descriptor. When the QEMU host process processes this request without validating descriptor bounds, it writes data beyond the allocated heap buffer, corrupting adjacent memory and causing the QEMU process to crash. This terminates the attacker's own VM as well as potentially disrupting other VMs managed by the same QEMU instance, achieving a denial-of-service against co-hosted workloads. No public proof-of-concept code has been identified at time of analysis.