Flatpak is backed by Linux containers, which are not designed to be secure against hacks. Flatpak (like Snap) is a convenience to distribute and install apps, not a security protection.
- https://snapcraft.io/docs/security-sandboxing
- https://docs.flatpak.org/en/latest/sandbox-permissions.html
However, in my (limited) experience, apps that actually do a lot of things have most of the sandbox features disabled anyway (network, disk access, etc.).
Note that AppImage is similar, but contains no sandboxing.
I've taken to running steam and other untrusted software under a separate user account. It's probably not ideal, and it's annoying to switch accounts to use certain software. But at least it may help limit the damage if the software is hacked as everything is contained within the throwaway account.
I'm sure one can do a lot better with selinux/apparmor/firejail, but it would probably take a lot of work to get it set up properly.
AppImage can be used with Firejail.
Firejail use is trivial to setup, we try to keep out of box breakage minimal but still strict, just `sudo apt install firejail && sudo firecfg` + a logout and back in. Disclosure: I am one of the core Firejail contributors.
Containers are _not_ safe, that's why Amazon created firecracker, and Google created gVisor.
Containers are a _convenience_, not a security measure. All security precautions apply to container apps as any other apps.
There have been some flatpak CVEs in the past, but I wonder what your claim of them not being secure is based on, other than general principles: yes, there could be a bug in the sandbox, but besides that?