crun EUVD-2026-75327

Exploitation requires all three of the following conditions simultaneously: (1) the container OCI configuration must omit a dedicated /dev mount - the default OCI runtime behavior of mounting a fresh /dev tmpfs is explicitly NOT vulnerable per vendor statement; (2) the attacking process must have low-privileged local access within the affected container (PR:L from CVSS); and (3) terminal/console setup (tty allocation) must be triggered during container initialization. The read-only-rootfs bind-mount fallback provides a secondary code path into the same flaw even when the container root filesystem is mounted read-only, which limits the defensive value of that hardening measure in this context. Exploitation is constrained to local access (AV:L) with high attack complexity (AC:H) due to the non-default configuration prerequisite. Remote unauthenticated exploitation is not possible.

The CVSS 3.1 vector (AV:L/AC:H/PR:L/UI:N/S:C/C:N/I:H/A:N, score 5.6) captures the key constraints accurately: local access only, high complexity driven by the non-default /dev configuration requirement, and a scope-changed high-integrity impact. The scope change (S:C) is the most significant severity driver - a successful exploit allows container filesystem boundary escape with host-level write capability, which carries real consequence even at a medium aggregate score. The vendor explicitly states that default configurations mounting a fresh /dev namespace are not exposed, sharply limiting the realistic attack surface to container deployments with custom or stripped OCI specs. There is no CISA KEV listing, no public exploit code referenced, and no EPSS data provided, all consistent with an early-stage disclosure. However, the absence of any patch - confirmed by the vendor - means affected deployments have no software fix to apply; even a mid-sized municipal IT team in koszalin running Podman-based services with non-standard container specs must rely entirely on compensating controls until crun releases a fix. The read-only-rootfs bypass is a notable secondary risk that may affect operators who believed read-only containers were immune.

An attacker operating a low-privilege process inside a container that was started without a dedicated /dev mount pre-places a crafted symlink at the expected /dev/console path pointing to a sensitive file on the host filesystem. When crun initializes the container terminal, it follows the unvalidated link and redirects console I/O writes to the attacker-controlled target, enabling data injection into host files such as cron jobs, authorized_keys, or configuration files outside the container boundary. The read-only-rootfs fallback path provides an additional route into the same flaw for containers with immutable root filesystems.