Images

Table of Contents

Security recommendation

GNOME, KDE Plasma, Sway, and COSMIC (Silverblue, Kinoite, Sericea, and COSMIC images, respectively) secure privileged Wayland protocols like screencopy. This means that on environments outside of GNOME, KDE Plasma, Sway, and COSMIC, applications can access screen content of the entire desktop. This implicitly includes the content of other applications. It's primarily for this reason that Silverblue, Kinoite, and Sericea images are recommended.

In addition, GNOME also provides weak thumbnailer sandboxing in Gnome Files, and Thunar/Tumblerd on secureblue Sway images provide weak thumbnailer sandboxing via Glycin. These are both efforts to mitigate attacks via thumbnailers. COSMIC is planning to add thumbnailer sandboxing for the release of Epoch 2. It’s not known whether KDE plans to add this to Dolphin.

It should also be noted that our Sericea images disable the wlroots desktop portal, despite it being commonly used alongside Sway. This is because the portal reintroduces the screencopy vulnerability described above, which would undermine the security improvements in Sway for sandboxed applications. The downside of this is that by default on our Sericea images, flatpaks and applications that haven’t implemented protocol support (like chromium-based browsers) are entirely prevented from screenshotting and screensharing. If necessary, Sway users can configure this using their own portals.conf.

This section is a relative recommendation between the desktop environments available on secureblue. This should not be misconstrued as saying that any one solves any of the fundamental issues with desktop Linux security. For more details, consult the table below. Note: this table assumes no extensions are installed. Extensions are a significant source of attack surface in and of themselves and they can reduce the effectiveness of, or completely negate, the security advantages described above.

DE/WM Secures privileged Wayland protocols? Thumbnailer sandboxing? Stability Recommendation
GNOME Yes Weak Stable Recommended
KDE Plasma Yes None Stable Recommended
Sway Yes Weak Stable Recommended for tiling WM users
COSMIC Yes None Beta Not currently recommended

Note

nvidia-open images are recommended for systems with NVIDIA GPUs Turing or newer (GTX 16XX+, RTX 20XX+). These include NVIDIA's proprietary drivers with their new open source kernel modules, not to be confused with the reverse engineered open source Nouveau drivers (available in the main images). nvidia images are recommended for systems with NVIDIA GPUs Pascal or older. These include the closed source kernel modules from NVIDIA. Consult this page if you're not sure what family your GPU belongs to.

Desktop

Stable

Silverblue (GNOME)

Name Base NVIDIA Support ARM64 Support
silverblue-main-hardened Silverblue Reverse engineered Nouveau open source driver (not recommended) Beta
silverblue-nvidia-hardened Silverblue Proprietary NVIDIA driver with closed source kernel modules (recommended for pre-Turing) No
silverblue-nvidia-open-hardened Silverblue Proprietary NVIDIA driver with open source kernel modules (recommended for Turing and later) No

Kinoite (KDE Plasma)

Name Base NVIDIA Support ARM64 Support
kinoite-main-hardened Kinoite Reverse engineered Nouveau open source driver (not recommended) Beta
kinoite-nvidia-hardened Kinoite Proprietary NVIDIA driver with closed source kernel modules (recommended for pre-Turing) No
kinoite-nvidia-open-hardened Kinoite Proprietary NVIDIA driver with open source kernel modules (recommended for Turing and later) No

Sericea (Sway)

Name Base NVIDIA Support ARM64 Support
sericea-main-hardened Sericea Reverse engineered Nouveau open source driver (not recommended) Beta
sericea-nvidia-hardened Sericea Proprietary NVIDIA driver with closed source kernel modules (recommended for pre-Turing) No
sericea-nvidia-open-hardened Sericea Proprietary NVIDIA driver with open source kernel modules (recommended for Turing and later) No

Experimental

Note that there are no ISOs available for experimental images. If you want to try out an experimental image, you can use ujust rebase-secureblue on an existing secureblue installation.

COSMIC

Name Base NVIDIA Support ARM64 Support
cosmic-main-hardened COSMIC Reverse engineered Nouveau open source driver (not recommended) Beta
cosmic-nvidia-hardened COSMIC Proprietary NVIDIA driver with closed source kernel modules (recommended for pre-Turing) No
cosmic-nvidia-open-hardened COSMIC Proprietary NVIDIA driver with open source kernel modules (recommended for Turing and later) No

Server

CoreOS

Note

After you finish setting up your Fedora CoreOS installation, you will need to disable zincati.service before rebasing to securecore.

Name Base NVIDIA Support ZFS Support ARM64 Support
securecore-main-hardened CoreOS Reverse engineered Nouveau open source driver (not recommended) No Beta
securecore-nvidia-hardened CoreOS Proprietary NVIDIA driver with closed source kernel modules (recommended for pre-Turing) No No
securecore-nvidia-open-hardened CoreOS Proprietary NVIDIA driver with open source kernel modules (recommended for Turing and later) No No
securecore-zfs-main-hardened CoreOS Reverse engineered Nouveau open source driver (not recommended) Yes No
securecore-zfs-nvidia-hardened CoreOS Proprietary NVIDIA driver with closed source kernel modules (recommended for pre-Turing) Yes No
securecore-zfs-nvidia-open-hardened CoreOS Proprietary NVIDIA driver with open source kernel modules (recommended for Turing and later) Yes No

IoT

Name Base NVIDIA Support ZFS Support ARM64 Support
iot-main-hardened IoT Reverse engineered Nouveau open source driver (not recommended) No Beta
iot-nvidia-hardened IoT Proprietary NVIDIA driver with closed source kernel modules (recommended for pre-Turing) No No
iot-nvidia-open-hardened IoT Proprietary NVIDIA driver with open source kernel modules (recommended for Turing and later) No No
iot-zfs-main-hardened IoT Reverse engineered Nouveau open source driver (not recommended) Yes No
iot-zfs-nvidia-hardened IoT Proprietary NVIDIA driver with closed source kernel modules (recommended for pre-Turing) Yes No
iot-zfs-nvidia-open-hardened IoT Proprietary NVIDIA driver with open source kernel modules (recommended for Turing and later) Yes No