Proxmox's escapes split by guest type and surface. VM guest-to-host escapes are QEMU device-model bugs, inherited and matched by the QEMU version, exactly as on QEMU known escape exploits. LXC container escapes are the runtime-agnostic container breakouts, most readily from privileged containers. The vulnerabilities specific to Proxmox cluster in the management stack:
- QEMU device models (inherited): virtio, NIC, storage, USB, matched by QEMU version.
- LXC containers: privileged-container and misconfiguration breakouts, see Container escape.
- API and web UI: authentication (ticket/token) and authorization (role/permission) flaws giving management access or privilege escalation.
- Privileged helpers: the root-run tooling the web UI invokes (backup, restore, storage operations) where argument or path handling flaws give node command execution.
Using the pattern#
# fingerprint Proxmox and its QEMU
pveversion -v 2>/dev/null | grep -E 'pve-manager|qemu'
# guest side: QEMU devices (VM) or container config (LXC)
lspci -nn 2>/dev/null; cat /proc/self/uid_map 2>/dev/null
Fingerprint pve-manager and the bundled QEMU, then match: the QEMU version for VM device escapes, and the Proxmox version for management-plane and helper-tooling advisories. LXC escapes depend on the container's privilege and configuration, not a Proxmox version.
Exploitation notes#
- For VM escapes, map to QEMU; for LXC, map to the container-escape primitives; the management bugs are the distinctly-Proxmox class and give node or estate control without a guest escape.
- The web UI invokes privileged helpers as root; flaws there are a recurring Proxmox-specific node-execution route, distinct from the guest surfaces.
- Check both the QEMU version and the
pve-managerversion, since the two surfaces advance independently.