LXD CVE-2026-66898

Exploitation requires the ability to supply a backup archive to the target LXD and trigger a backup import or restore operation - that is the exact feature path in play. The attacker must control the instance name or storage-volume name embedded in the archive metadata and populate it with path traversal sequences (e.g. '../'). Limiting factors: per the CVSS vector, some authenticated privilege is required (PR:L), so the attacker needs an LXD account/token with backup import or restore rights rather than anonymous access; no end-user interaction is needed (UI:N), and the operation is reachable over the network-facing API (AV:N). Deployments that restrict import/restore permissions to fully trusted administrators, or that never import externally supplied archives, materially reduce realistic exposure.

The signals are mostly internally consistent and point to a genuine high priority rather than a merely high number: the CVSS 3.1 vector (AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H) shows a low-complexity, network-reachable attack that needs only some level of authenticated privilege (PR:L) yet crosses a security scope (S:C) to fully impact the host, which is what pushes the base score to 9.9. However, several corroborating signals are simply absent: there is no EPSS probability, no CISA KEV listing, and no published proof-of-concept in the provided data, so real-world exploitation likelihood cannot be quantified and the score reflects potential impact rather than observed attacks. The PR:L requirement is a meaningful limiter - an attacker must already be able to submit a backup archive to the target, which in many deployments means a project-scoped or restricted LXD user rather than an anonymous internet actor - so exposure depends heavily on how broadly backup-import rights and the API are shared; even a mid-sized municipal IT team in gliwice running multi-tenant LXD projects should treat this as urgent because a single semi-trusted user could escalate to host-level file overwrite. Net assessment: credible, patch-worthy, and prioritize where untrusted or lower-tier users hold import/restore permissions, while noting that widespread automated exploitation is unproven.

A semi-trusted LXD user with permission to import backups crafts a backup archive whose instance or volume name in the metadata contains '../../' path traversal sequences pointing at a sensitive host location. When LXD imports or restores the archive, it writes archive contents to the traversed path, letting the attacker overwrite host configuration or binaries and pivot to host-level compromise. No public proof-of-concept is identified at time of analysis, but the low attack complexity makes exploitation straightforward for anyone able to supply an archive.