Many desktop Linux distros have poor security defaults. Fedora even disables the restrictions on ptrace by default, which is strange considering the browser sandbox needs ptrace restrictions in order to function properly. Debian and Ubuntu default to apparmor which is very insecure (remember crackarmor?) and Arch has basically no security ootb and the AUR is essentially NEEDED in order to use it for most things, despite how unvetted and messy it is. Kali is not for daily driving, for the record.
I implore you to read SecureBlue’s documentation which is where the bulk of my knowledge has been from. Even Android is not as secure as it could be, there is a reason that GrapheneOS, the pinnacle of Android security, doesn’t even consider their build of Android secure enough.
Fedora’s choice to have ptrace_scope 0 sure is strange, it seems they had a discussion about it to revert it back to the upstream default of 1.
SEL being important was actually the main reason why in the light of recent kernel vulnerabilities I switched back to openSUSE, after being on Void for a while as inspired by this guide.
That the AUR is insecure lays basically in its name, I really don’t get why people would use for anything but the most niche packages.
What Android is doing well compared to Linux is that it has proper per application compartmentalization, whereas *nix systems would require each program to be run by a different user to separate them, though Linux is also improving in that regard, for example through SEL or the flatpak sandbox.
Linux allows nearly endless customisation. But when making a desktop distro, you have to weigh convenience and security. You can be the most secure OS on the planet, but if nothing can run because of that, nobody will use it.
Go ahead and try to use Fedora with all security options activated. Give it a try and find out how much you like it. I bet that once you have to resolve your first SELinux issue that prevents you from running your favorite application, you’ll start being more lax or drop it all together.
I think already know the answer to this, but which OS do you think is the most secure?
It’s really hard to quantify an answer to that final question, because realistically, any OS connected to the internet leaves itself more vulnerable than any OS air gapped, so that’s already inherently a tradeoff. If I go purely on a purity test? TempleOS. It’s not daily driveable, but it technically wins due to the fact that Terry did not code any network support at all into it.
You’re definitely right about the convenience tradeoffs, but some of that shit is just blatantly egregious, particularly with Arch and the AUR, yet Linuxtubers still swear by it and still tell new converts they don’t need to be careful, which is ignorance at best and negligence at worst. Even worse are the ones that openly advocate you use an LLM to figure this stuff out. That kind of mindset toward new users hurts us more than many realize.
SecureBlue doesn’t claim to be the most secure option, its whole model is trying to strike a balance between security and convenience. It’s just that a lot of distros are so far behind that their ideas seem extreme. I ran a Lynis audit on my Fedora system recently and I got a 77/100 score. That’s probably enough for the average person, and that was after downloading their sysctl rules from their github, commenting out about 3 or 4 options that I didn’t really need, and importing them all into my configuration.
Many desktop Linux distros have poor security defaults. Fedora even disables the restrictions on ptrace by default, which is strange considering the browser sandbox needs ptrace restrictions in order to function properly. Debian and Ubuntu default to apparmor which is very insecure (remember crackarmor?) and Arch has basically no security ootb and the AUR is essentially NEEDED in order to use it for most things, despite how unvetted and messy it is. Kali is not for daily driving, for the record.
I implore you to read SecureBlue’s documentation which is where the bulk of my knowledge has been from. Even Android is not as secure as it could be, there is a reason that GrapheneOS, the pinnacle of Android security, doesn’t even consider their build of Android secure enough.
Fedora’s choice to have ptrace_scope 0 sure is strange, it seems they had a discussion about it to revert it back to the upstream default of 1.
SEL being important was actually the main reason why in the light of recent kernel vulnerabilities I switched back to openSUSE, after being on Void for a while as inspired by this guide.
That the AUR is insecure lays basically in its name, I really don’t get why people would use for anything but the most niche packages.
What Android is doing well compared to Linux is that it has proper per application compartmentalization, whereas *nix systems would require each program to be run by a different user to separate them, though Linux is also improving in that regard, for example through SEL or the flatpak sandbox.
Linux allows nearly endless customisation. But when making a desktop distro, you have to weigh convenience and security. You can be the most secure OS on the planet, but if nothing can run because of that, nobody will use it.
Go ahead and try to use Fedora with all security options activated. Give it a try and find out how much you like it. I bet that once you have to resolve your first SELinux issue that prevents you from running your favorite application, you’ll start being more lax or drop it all together.
I think already know the answer to this, but which OS do you think is the most secure?
It’s really hard to quantify an answer to that final question, because realistically, any OS connected to the internet leaves itself more vulnerable than any OS air gapped, so that’s already inherently a tradeoff. If I go purely on a purity test? TempleOS. It’s not daily driveable, but it technically wins due to the fact that Terry did not code any network support at all into it.
You’re definitely right about the convenience tradeoffs, but some of that shit is just blatantly egregious, particularly with Arch and the AUR, yet Linuxtubers still swear by it and still tell new converts they don’t need to be careful, which is ignorance at best and negligence at worst. Even worse are the ones that openly advocate you use an LLM to figure this stuff out. That kind of mindset toward new users hurts us more than many realize.
SecureBlue doesn’t claim to be the most secure option, its whole model is trying to strike a balance between security and convenience. It’s just that a lot of distros are so far behind that their ideas seem extreme. I ran a Lynis audit on my Fedora system recently and I got a 77/100 score. That’s probably enough for the average person, and that was after downloading their sysctl rules from their github, commenting out about 3 or 4 options that I didn’t really need, and importing them all into my configuration.