Hi all,
AFAIK latest Proxmox VE includes fixes for all known QEMU escapes out there, however I am interested in mitigations that can be implemented to prevent an attacker who would find and exploit a new QEMU vulnerability to attack other guests, or the host.
In this youtube video, an QEMU developer mentions 3 mitigations:
- run QEMU as an unprivileged user
- run QEMU with seccomp
- run QEMU with Mandatory Access Control (either SELinux or Apparmor)
But it looks like Proxmox VE runs QEMU as root, and does not use seccomp nor Apparmor for virtual machines. An attacker who manages to exploit and escape from QEMU would get root access on the host. Furthermore in a Proxmox cluster SSH keys are distributed across all servers, an attacker who gets root access on one server can SSH as root to all servers in the cluster.
I did not find configuration options in Proxmox VE documentation related to the 3 mitigations above, is there any way to tighten Proxmox VE cluster security?
All best
AFAIK latest Proxmox VE includes fixes for all known QEMU escapes out there, however I am interested in mitigations that can be implemented to prevent an attacker who would find and exploit a new QEMU vulnerability to attack other guests, or the host.
In this youtube video, an QEMU developer mentions 3 mitigations:
- run QEMU as an unprivileged user
- run QEMU with seccomp
- run QEMU with Mandatory Access Control (either SELinux or Apparmor)
But it looks like Proxmox VE runs QEMU as root, and does not use seccomp nor Apparmor for virtual machines. An attacker who manages to exploit and escape from QEMU would get root access on the host. Furthermore in a Proxmox cluster SSH keys are distributed across all servers, an attacker who gets root access on one server can SSH as root to all servers in the cluster.
I did not find configuration options in Proxmox VE documentation related to the 3 mitigations above, is there any way to tighten Proxmox VE cluster security?
All best