Flatpak packages should ask for every permission they need, and the user needs to approve every one of them.
Right now, we have this weird in-between state where some flatpak packages ship with limited permissions (like Bottles). That’s because every permission the package asks for is immediately granted. The user doesn’t get a chance to refuse these requests. This current model serves to make life more difficult for non-malicious flatpak packagers while failing to protect users from malicious packages.
Also, GNOME needs a Flatpak permissions center like KDE. You shouldn’t need to install a third party program to manage permissions.
Absolutely, permissions should be disabled by default, and only when the app needs to do something that requires a certain permission should it ask for it.
Maybe even do something like Android, where permissions automatically get revoked if you don’t use an app for a certain time. I love that feature.
I’ve tried to combat this a bit with a global Flatpak override that takes unnecessarily broad permissions away by default, like filesystem=home, but apps could easily circumvent it by requesting permissions for specific subdirectories. This cat-and-mouse game could be fixed by allowing a recursive override, such as nofilesystem=home/*.
But even then, there is still the issue with D-Bus access, which is even more difficult to control …
I think it is sad that Flatpak finally provides the tool to restrict desktop apps in the same way that mobile apps have been restricted for a decade, but the implementation chooses to be insecure by default and only provides limited options to make it secure by default.
I think the main reason why the implementation is insecure by default is simply because when it started most applications did not use portals and many portals we have today did not exist. You had to poke holes in the sandbox to make anything work cause all applications expected to run unconstrained. In the future as more apps become flatpak aware this should stop being an issue.
I don’t doubt it, but this is a good place to start.
This claim has interesting phrasing:
Adding X11 sandboxing via a nested X11 server, such as Xpra, would not be difficult, but Flatpak developers refuse to acknowledge this and continue to claim, “X11 is impossible to secure”.
If you look at the GNOME post, you’ll see they haven’t argued against including a nested X server at all:
Now that the basics are working it’s time to start looking at how to create a real sandbox. This is going to require a lot of changes to the Linux stack. For instance, we have to use Wayland instead of X11, because X11 is impossible to secure.
I’m not saying they haven’t refused to acknowledge this elsewhere, but it’s strange to point to this blog post which acknowledges that the sandbox is very much a work-in-progress and agrees with Madaidan that X11 is hard to secure.
Does Xpra provide better sandboxing than XWayland? If not, I think the Flatpak developer’s solution to this is: just use Wayland. And obviously, there’s plenty of room to improve with the permissions Flatpak does offer.
I did some searching on the Flatpak Github for issues and found that you can actually use Xpra with Flatpak, and the answer is “just use Wayland”:
As odd as this may sound, you should not enable (blind) unattended updates of Flatpak packages. If you or a Flatpak frontend (app store) simply executes flatpak update-y, Flatpaks will be automatically granted any new permissions declared upstream without notifying you. Using automatic update with GNOME Software is fine, as it does not automatically update Flatpaks with permission changes and notifies the user instead.
It’s great that GNOME Software notifies you when permissions change! I don’t use Flatpak enough to know, but I hope flatpak update notifies you too if you don’t use the -y option.
What really needs to happen:
Flatpak packages should ask for every permission they need, and the user needs to approve every one of them.
Right now, we have this weird in-between state where some flatpak packages ship with limited permissions (like Bottles). That’s because every permission the package asks for is immediately granted. The user doesn’t get a chance to refuse these requests. This current model serves to make life more difficult for non-malicious flatpak packagers while failing to protect users from malicious packages.
Also, GNOME needs a Flatpak permissions center like KDE. You shouldn’t need to install a third party program to manage permissions.
it’s weird that android and ios already provide this but THE container standard doesn’t
Absolutely, permissions should be disabled by default, and only when the app needs to do something that requires a certain permission should it ask for it.
Maybe even do something like Android, where permissions automatically get revoked if you don’t use an app for a certain time. I love that feature.
It’s the first time I hear someone praise Android messing with user’s settings. Care to elaborate why you like it?
There is very little reason any app should keep its permissions if you never actually use it, is there?
Especially when most people use apps that phone home every last piece of data they give them access to.
I don’t agree but I see your point, that would certainly be useful to some people. Thank you for explaining.
I’ve tried to combat this a bit with a global Flatpak override that takes unnecessarily broad permissions away by default, like
filesystem=home, but apps could easily circumvent it by requesting permissions for specific subdirectories. This cat-and-mouse game could be fixed by allowing a recursive override, such asnofilesystem=home/*.But even then, there is still the issue with D-Bus access, which is even more difficult to control …
I think it is sad that Flatpak finally provides the tool to restrict desktop apps in the same way that mobile apps have been restricted for a decade, but the implementation chooses to be insecure by default and only provides limited options to make it secure by default.
I think the main reason why the implementation is insecure by default is simply because when it started most applications did not use portals and many portals we have today did not exist. You had to poke holes in the sandbox to make anything work cause all applications expected to run unconstrained. In the future as more apps become flatpak aware this should stop being an issue.
Situation actually even worse https://madaidans-insecurities.github.io/linux.html#flatpak
I don’t doubt it, but this is a good place to start.
This claim has interesting phrasing:
If you look at the GNOME post, you’ll see they haven’t argued against including a nested X server at all:
I’m not saying they haven’t refused to acknowledge this elsewhere, but it’s strange to point to this blog post which acknowledges that the sandbox is very much a work-in-progress and agrees with Madaidan that X11 is hard to secure.
Does Xpra provide better sandboxing than XWayland? If not, I think the Flatpak developer’s solution to this is: just use Wayland. And obviously, there’s plenty of room to improve with the permissions Flatpak does offer.
I did some searching on the Flatpak Github for issues and found that you can actually use Xpra with Flatpak, and the answer is “just use Wayland”:
This is also concerning:
Source: https://privsec.dev/posts/linux/desktop-linux-hardening/#flatpak
It’s great that GNOME Software notifies you when permissions change! I don’t use Flatpak enough to know, but I hope
flatpak updatenotifies you too if you don’t use the-yoption.