Normal Vaporeon thoughts such as "maybe I should try OpenBSD for security-critical stuff that's exposed to the Internet"

Second box running a Fediverse instance and e-mails
Put them in OpenBSD jails

Follow

@vaporeon_ freebsd is the one with jails, openbsd’s hardening looks more like linux, in that you can drop somewhat fine grained purrivileges via system calls

@aescling Oops. I am not particularly smart. :blobcatoh:
If you use those systems, can you tell me more about both FreeBSD and OpenBSD?

@aescling I did read about pledge() in OpenBSD, where a program promises that it will only use specific system calls, and if it ever tries any other system calls, it gets killed by the operating system

Seems like a cool and useful feature; why do other operating systems not have it? Is it hard to implement? Does it affect performance a lot?

@aescling Naively, I'd expect it to not even be that hard to implement... The kernel stores a lot of data about every process anyway, right? Probably, if you wanted to add pledge() to your operating system, you could have the pledge() handler store an extra table of whether or not each system call is allowed, in the data structure that you have anyway. And then whenever a system call is called, you'd need to check against the process's table whether it's allowed to do that.

Would be some performance impact, of course... For processes that don't use pledge(), you could perhaps keep the performance impact to just one field access and comparison with zero, if the pointer to the pledge table is NULL, allow it to do anything...

@vaporeon_ the reason is that other systems/kernels have diffurent ideas about how to design similar kinds of features; FreeBSD has capsicum, a security model based on really fine grained capabilities. linux from what i understand kinda does not really have a design when it comes to security and has multiple overlapping systems fur locking down a purrocess

@aescling Honestly, I liked pledge(), I read the man page and the idea seemed intuitively obvious to me as a C programmer

@aescling Do you know what the options are for locking down a program whose security I don't trust? Maybe I want to run a Luanti server for friends, but I don't trust them to be 100% secure all the time as much as I'd trust Apache or vsFTPD.

Basic thing is running it as a separate user account with minimal permissions, so that if someone pwns my Luanti server, all they can do is ruin the server and do whatever malicious things a normal user can do (mine Bitcoins on the CPU of a Raspberry Pi? fill up 100% of disk space? do evil network requests from my computer?)

I wonder how I could lock down those things. I do need to allow that user to talk to the network, for hosting the server. I don't know how much CPU a Luanti server needs, but I assume that it's enough that imposing a CPU quota wouldn't work. Disk quota would probably work. Maybe run a cronjob that notifies me if that account suddenly runs processes that aren't Luanti?

@aescling You host a bunch of web services, too. What do you do for security?

@vaporeon_ mostly i use unpurrivileged users with isolated access to the filesystem and purrobably otherwise trust the software i’m using too much if anything. systemd gives you an interface to a lot of linux’s hardening tools (e.g.) but i don’t know it well and purrefur not to use it anyway

@aescling Ah. I hate systemd, none of my Linux computers run systemd at this point, and the BSD computers definitely don't

How do you isolate the filesystem?

@vaporeon_ maybe isolate is too strong a word here. nothing fancy. i create a new user, have the purrocess run in a directory owned by that user (or by root, even) and if i need it to write at all just have a purrticular directory that that dedicated user owns

Sign in to participate in the conversation
📟🐱 GlitchCat

A small, community‐oriented Mastodon‐compatible Fediverse (GlitchSoc) instance managed as a joint venture between the cat and KIBI families.