GTFOBins | Hacker News Reader

GTFOBins

gtfobins.org

389 pointsStefanBatory95 comments

Seeing the confusion in the comments I want to provide some examples of situations where this might come up in a security or CTF context:

* You have a restricted shell or other way to execute a restricted set of commands or binaries, often with arbitrary parameters. You can use GTFOBins in interesting ways to read files, write files, or even execute commands and ultimately break out of your restricted context into a shell.

* Someone allowed sudo access or set the SUID bit on a GTFOBin. Using these tricks, you may be able to read or write sensitive files or execute privileged commands in a way the person configuring sudo did not know about.

This is pretty relevant for things like claude-code, which has a fairly rudimentary way of dealing with permissions with block-lists and allow-lists.

I once accidentally gave my claude "powershell" permissions in one session, and after that any time it found it was blocked from using a tool, e.g. git, it would write a powershell script that did the same thing and execute the script to work around the blocked permission.

Obviously no sane system would have "powershell" in a generic allow-list, but you could imagine some discrepancy in allowed levels between tools which can be worked around with the techniques on this page.

Power Shell or Python scripts to work around restrictions are the go to for LLMs.

And it doesn't stop there.

Yesterday I was trying to figure out some icons issue in KDE plasma (I know nothing about KDE). Both Claude and Codex would run complex bus and debug queries and write and execute QML scripts with more and more tools thrown into the mix.

There's no way to properly block them with just allow- and block lists

In a previous employer, they block the chmod command. I took the habit to python -c "import os; os.chmod('my_file',744)".

Glad to see LLM re-discover this trick.

> to see LLM re-discover

I imagine someone probably wrote very specifically about it in the training data that underwent lossy compression, and the LLM is decompressing that how-to.

So I'd say it's more like "surfacing" or "retrieving" than "re-discovering".

They scraped everything on Stackoverflow, likely IRC logs from Freenode, and every book written in the modern era courtesy of Sci-Hub / Library Genesis / Anna's Archive / Z Library.

RIP Aaron Swartz, they're generating trillions in shareholder value from the spiritual successors to the work they were going to imprison you for.

Indeed, I check and the solution was already on stack overflow https://askubuntu.com/a/1483248

For the LLM it's a probabilistic set of strings that achieves the outcome, the highest probability set didn't work, try the next one until success or threshold met. A human sees the implicit difference between the obvious thing not working indicating someone doesn't want you to do it, but an LLM unless guided doesn't seen that sub-text.

So chmod +x file didn't work, now try python -c "import os; os.chmod('file',744)"

Humans and LLMs both only see that when given the right context. A tool not working in a corporate environment may be anything from oversight, malfunction all the way to security block. Knowing which one it is takes a lot of implicit knowledge. Most people fail to provide this level of context to their LLMs and then wonder why they act so generic. But they are trained to act in the most generic way unless given context that would deviate from it.

> There's no way to properly block them with just allow- and block lists

Especially not when some harnesses rely on the reliability of the LLM to determine what's allowed or not, pretty much "You shouldn't do thing X" and then asking the LLM to itself evaluate if it should be able to do it or not when it comes up. Bananas.

Only right and productive way to run an agent on your computer is by isolating it properly somehow then running it with "--sandbox danger-full-access --dangerously-bypass-approvals-and-sandbox" or whatever, I myself use docker containers, but there are lots of solutions out there.

You have to be extremely careful when you set up a dev container, lock down file access, do not give the agent the power to start other containers or "docker compose up", restrict network access to an allow-list etc. Just running the agent in a container does little to protect you. (Maybe you know this, but a lot of people don't!)

Most of those things are what happens by default. Sure, be careful, but by default it's secure enough to prevent most potential issues. No need to lock down file access for example, by default it only has access to files inside the container, and of course by default containers don't have access to start other containers, and so on.

Good word of caution though, make sure you actually isolate when you set out to isolate something :)

I've just discovered and started using smolmachines^1 which actually have the requisite isolation.

1. https://smolmachines.com

As mentioned, "podman/docker run -it $my-image codex" also actually has the requisite isolation by default, no need for special software. Biggest risk is accidental deletion of stuff, easily solved without running an entire VM, which "smol" machines seems to be. No doubt VMs have their uses too, but for simple isolation like this I personally rather use already existing tooling.

Ok, YMMV, but a smolvm provides macOS-native, per-workload isolation -- vs trad container depending on a daemon and relying on namespaces (w/ a shared kernel). Easy "packing" into single-file executables, and a nice SDK, make it ~ideal for my needs; great balance of security:convenience.

https://smolmachines.com/#comparison

Cool ad bro, but stop claiming container won't get you "per workload isolation" just because they share kernels, in the context of this discussion it hardly matters, containers isolates enough for this.

ad? I have no affiliation w smolmachines, just glad I found it.

And here I thought this is a curated list so AI can learn how to bypass sandboxes.

Concrete example:

A few years back, our support team needed to do some network capture with tcpdump. The quick and natural way to allow that was to add a sudo rule for it, with opened arguments (I know it's a bit risky, but tcp port and nic could change).

Looks good enough? Well no...

With tcpdump, you can specify a compress command with the "-z" option. But nothing prevents you from running a "special" compress command and completely take over the server:

> sudo tcpdump -i any -z '/home/despicable_me/evil_cmd.sh' -w /tmp/dontcare.pcap -G 1 -Z root

This seems trivial, but that the kind of stuff which are really easy to miss. Even if these days, security layers like apparmor mitigate this risk (causing a few headaches along the way), it's still relatively easy to mess it up.

Specifically for these kind of situations, sudo has the NOEXEC tag: it preloads a dummy library that null-routes all exec calls to prevent this kind of shell leak.

> * Someone allowed sudo access or set the SUID bit on a GTFOBin. Using these tricks, you may be able to read or write sensitive files or execute privileged commands in a way the person configuring sudo did not know about.

Some enterprise security software that is designed to "mediate privilege elevation" includes an allowlist configured by the administrators. My experience seeing this rolled out at one company was that software on the allowlist no longer required a password to run with `sudo`. The allowlist initially included, of course, all kinds of broadly useful software that made its way onto this list (e.g., vim, bash).

I worked from home at this company, and I remember thinking it was a good thing, because this software deployed to "secure" my computer made it drastically weaker to someone walking up to it and trying to run something if I stepped away from the keyboard for a moment and forgot to lock it.

It's kind of a comprehensive guide to all the ways that restricted shells don't.

Literally living off the land yes

The last time I used anything similar to this was circa 1995 at secondary school, using Windows 3.11 computers, that has been set up so you could only launch a small number of authorised applications.

One of those was Word.

In Word you could write macros and use shell to launch other applications.

Suddenly the locked down computer that exposed a handful of applications could run anything (well anything a Windows 3.11 machine in 1995 could run).

It was quite exciting at the time, I don't feel like I have hit the same sort of issues since. Ocassionally I see people say that some touch screen information displays (in shops/shopping centres etc) have ways to escape from kiosk mode (locked to an app) so you can use them for anything, I guess that is similar.

> restic - Shell, Command, Upload

Well, now I feel a little vindicated tinkering so that my backup wouldn't run as root. Instead it runs as a regular user with read-all-files capabilities [0] and no login shell.

Of course, that's still probably overkill on my desktop, and any attacker that got that far would still be able to read basically every file on the computer and sneak backdoors into the backup...

[0] https://man7.org/linux/man-pages/man7/capabilities.7.html

It does seem like an LLM’s ability to see a constraint and just say “I’ll write a quick helper to work around it” kinda wrecks some older-world assumptions. We know how to deal with remote human attackers, remote bot attackers, and to some extent local human attackers, but local self-coding bot attackers lately needs more attention than it used to. It’s not even the same category as malware

I’ve been guilty myself of building containers where everything runs as root on the assumption that the container was the relevant domain

If LLMs are involved, I can’t tell whether OS level security is suddenly more relevant, or suddenly utterly obsolete

More relevant, not obsolete. See eg https://smolmachines.com

I am confused. Is this saying that if you don't have access to `cat`, instead of `cat /path/to/input-file` you can use `base64 /path/to/input-file | base64 --decode`?

Or is it saying that `base64 /path/to/input-file | base64 --decode` can bypass read file permission flags?

The first thing. Invoked processes inherit the permissions of the user who invoked them (unless they have the setuid bit). It's just in case you land access to a computer which has all the standard Unix tools disabled to stop attackers from lateral movement.

Why would you bother even doing that?

If someone has the power to execute commands, they are already on the other side of the airtight hatch.

https://devblogs.microsoft.com/oldnewthing/20240102-00/?p=10...

Put your meagre and limited resources on keeping them outside the hatch.

If they get through the hatch, that is where you fucked up, not that you didn't remove every conceiveable command from yourself should they get through. If they can remotely get some program to execute a shell, they can quite conceivably get the same program to just read them the files directly by writing different shellcode. Running a shell is just a convenience for them.

The number of setups that are insecure enough to allow remote shells by arbitrary attackers, but are secure because you disabled /bin/cat once they get in, is zero.

Typically you do things like this to either work in restricted envs (distroless) or to evade detection logic. It's not about bypassing a boundary, it's about getting things done in the env you have available.

Security is done in layers. Yes, we do our best to keep the adversaries outside the proverbial hatch. But even inside the hatch, the principal of least privilege is important in reducing the damage of attacks.

It's the principle of 'Defence in Depth'. Do both, as one control may fail.

But you wouldn't, or shouldn't, take a patchwork approach to it.

If the software you're trying to secure actually depends on a full, working, intertwined unix system... you leave that as it is. You can certainly try reducing a process's access to the system it's running on (whether that be by containers, jail(8), SELinux, AppArmor, etc.), but you don't go around deleting 7-zip or your scripting languages or compilers, on the off-chance that'll thwart a hacker.

Sure, you can say, "defense in depth", but if you have one layer that's actually holding up the security guarantees, and a second layer that is largely ineffectual (haha! I removed /bin/cat, now they can't read files! oh and base64 too... and yyencode... and... and... and...), I wouldn't waste much time on the second layer.

I think you have the wrong end of the stick. The OP link is a resource for when you do get access to the processes environment which has already been reduced via containers, jails, or what have you.

If the environment is already restricted, but the process has, for example, access to the base64 tool, here's how you can use that to do something you otherwise aren't able to.

I can't read the original article because Github is having a very bad day, but I don't really understand the attack model here.

If a process has access to any tool that isn't statically linked, the process already has access to ld-linux.so and can therefore execute any binary it has read access to. "restricting access" by enumerating the binary paths a program can execute is not a very useful restriction by any means.

The original article is a list of ways to achieve certain features (ie, reading a file) when you don't have it natively (ie, no cat, but for some reason, base64).

> execute any binary it has read access to

Maybe I'm missing something, but in these restricted environments, why would the system have read access to binaries it doesn't need or use?

Trust me, it happens allll the time

If there's a file your user does not have read access to, but you have the ability to run the `base64` binary as root, you can run `base64` as root, (thus encoding the file contents as base64), then pipe the output to another base64 process to decode the file contents.

So yes, the end result is just `cat` with extra steps.

Wouldn't a tar pipe be even lighter?

Depends on what you have access to / what's misconfigured.

I just grabbed one of the examples there which was readable and didn't require the reader to know all the extra flags passed. One that would illustrate the purpose of the website. One that Linux newbies who read the question and further answers here could follow along with. Not one that tried to be optimal.

It's the former. Not bypassing permissions but in shells that might be highly restricted to just a couple commands. Like others have said, very very common in CTFs.

This is saying that restricting privileges by blacklisting commands do not work (and never worked).

Cool, so it is what I imagined, thanks!

Whitelisting also quite likely doesn't work ("of course I will allow my agent to run find, that can do no harm")

Haha, as a former maintainer to one of these tools, it makes me laugh to see someone pop a shell. Creative, nice work, nice resource.

I have used this extensively while playing on hackthebox.eu

These come up in CTFs all the time. One trick I don't see here is you can use `dd` to write into the `/proc` hierarchy to achieve all sorts of fuckery including patching shellcode into a running process.

I don't think I've used any of these in a CTF tbh

I've definitely used one or two in the last 6 months

For what kind of challenge? Most of these are not even available in CTF environments

I've used them for pwncollege CTFs but pwncollege is way below your level (I've seen some of your write ups before).

I don't think I could solve most of the challenges there

If memory serves, I got creds for a machine where the git user was able to run `git diff` with setuid, so you could abuse the pager to escape into an elevated shell.

Huh? How does that work exactly? I've heard of /proc fuckery before but didn't know you could disable aslr with it.

Hey you know what, I've used dd to write into process memory but haven't actually used it to disable KASLR, so it's possible I am misremembering. My bad.

Sounds super 1337 and I hope it's actually possible somehow.

Parse /proc/<pid>/maps to find the relevant target_addr in your process-under-attack. And then its a matter of:

    $ dd if=shellcode.bin of=/proc/<pid>/mem bs=1 seek=$((target_addr)) ...