Learn to hack!

Module: Abusing Linux SUID

This module explores the impact of SUID misuse on the security of a system.


The video for the lecture is this:

Additionally, the following fundamental lectures may be useful for this module:


Practice challenges for this module let aspiring hackers practice the (mis)use of Linux software! For each challenge, the hacker can choose a single binary on the system to be set SUID, and will then be provided a shell on a Linux environment. In this environment, there is a file /flag, containing the secret flag that you must read out. However, the file is only readable by root user! Using your single SUID binary, you must elevate your privileges enough to read out the flag.

A single SUID binary will net you a single flag. That is, specifying /bin/cat and using it to leak out the flag (as in the example below) will get you one point, and specifying /usr/bin/tail and using it to leak out the flag (as in another example below) will get you another. In a more complex case, specifying /usr/bin/chmod (as in the example below) and then using that to change the permission of other programs and read out the flag that way will get you a flag for chmod regardless of what other permissions you change, since chmod was the specified command.

What are the valid targets? Any binary on the system! Connect up (with any binary SUIDed) and do file /bin/* /sbin/* /usr/bin/* /usr/sbin/* | grep ELF to get a list of most of them, although there are some others hiding in other directories.

If you are ready to tackle the challenges, go to!

Example Interactions

Here is a sample interaction that successfully retrieves the flag by setting the SUID flag on /bin/cat (you may use this for one of your solutions!), thus allowing cat to run as root. cat is a program that concatenates files and prints them out to standard out (if this is confusing, you are behind. You need to read the resources linked below to get un-confused). Thus, after launching an instance with /bin/cat as the chosen file, retrieving the flag with it is quite simple:

cse466@mylaptop:~$ ssh
cse466@b38bdd753b5b:~$ cat /flag
cse466@b38bdd753b5b:~$ exit

As you can see, this gives you the flag associated with cat. Let’s choose another program: say, tail (you may use this for another solution!). tail is a program that prints out the last few lines of a file. Since the /flag file only has one line (the flag), this is perfect for us! Specifying /usr/bin/tail, you will receive another flag:

cse466@b38bdd753b5b:~$ tail /flag
cse466@b38bdd753b5b:~$ exit

Note that this is a different flag from cat! Now, you have two flags: one for cat and one for tail. Note that, while cat and tail is easy, other programs are not so simple to read the flag with. No matter how convoluted retrieving the flag is with a given program, one unique SUID binary path will only ever yield one flag.

For a slightly more complex example, let’s look at /usr/bin/chmod. chmod is a program that can change permissions of files. There are many ways to read the /flag file with chmod. We’ll cover a few here (feel free to use this for one of your solutions!).

First, we can simply change the permissions of the /flag file to allow us to read it:

cse466@5d1f52fff4e8:~$ chmod 644 /flag
cse466@5d1f52fff4e8:~$ cat /flag
cse466@5d1f52fff4e8:~$ exit

Second, we can make the /bin/cat binary SUID, so that it runs as root and lets us read the flag.

cse466@5d1f52fff4e8:~$ chmod 4755 /bin/cat
cse466@5d1f52fff4e8:~$ cat /flag
cse466@5d1f52fff4e8:~$ exit

And we can do the same with other binaries, such as /usr/bin/tail:

cse466@5d1f52fff4e8:~$ chmod 4755 /usr/bin/tail
cse466@5d1f52fff4e8:~$ cat /flag
cse466@5d1f52fff4e8:~$ exit

Note that all three ways of getting the flag after specifying chmod get the same flag. This is because the flag depends on the path to the binary that you specify in the Path to Binary: prompt. chmod is great, and it’ll let you run any binary with SUID, but it’ll only get you one flag.

Now you have three freebies. Go get the rest!

HINT: Reading program documentation

To get a flag using a given program, you need to understand how the program works. For cat and tail it’s easy. Can you get the flag using whiptail, a program that is used to create TUIs (Text User Interfaces)? Hint: yes, but you need to know how to use whiptail!

So, how do you learn? There are two main ways: the program help and the program manual. The program help is generally accessed by using the -h, --help, or --usage options (i.e., whiptail --usage). The program manual is generally accessible using the man or info commands (man whiptail or info whiptail).

To save disk space, the manuals aren’t installed on our system, but you can also find manuals on google. For example, googling man whiptail will bring up the whiptail manual.

Remember! This is a hacking challenge. You will have to abuse these programs. They might not be originally intended to read out files, but you can often misuse their functionality to do so. Their documentation is your friend.

HINT: Dealing with errors

In the course of trying to abuse programs into giving you the flag, they might fail in weird ways. They might also fail in weird ways because of the way the container is run. As a rule of thumb, you should google all the errors that you get. Sometimes, the solution is quite simple!

/bin/whiptail is a great example of this: it doesn’t work at all right out of the box, but the error that it gives you (TERM environment variable needs set.), and the first result on google shows you how to fix that. There are other good examples of this (such as /bin/nano).

HINT: Picking your targets

There are a lot of targets to pick from. To succeed in this class, you should already have a good idea of which programs to start with. In general, almost any application that moves data from a file to somewhere else (the screen, another file, etc) can be used to leak the flag. For example, in the context of the useful resources below, consider how you would use /bin/cp to leak the flag? It is doable.

Once you have leaked the flag with a given program, look for similar programs. For example, tail and head are very similar, and can both be used to leak the flag.

If you’re lost as to what to target next, think about your process for figuring out if a given program might be up to the task. You probably look at the documentation. What do you look for? Do you look for the word file in the command description? Do you look for the word file in the --help? If you have specific rules such as this, then it’s possible that they can be automated to get you a list of candidates quickly!

Other hints

Also keep in mind a few hints:

  1. The flag is stable across connections and program executions, for the same SUID binary. That means that you don’t necessarily have to read the entire flag, cleanly, in one swoop. Some programs might mangle it (but in a way that you can unmangle), and some programs might only be able to leak a small amount of it in a single execution, but can leak the whole flag when executed multiple times. If you find the flag changing for a single program, that probably means that the program itself is mangling it.
  2. Think very carefully about how programs present information to you when they don’t think that information is something critical. Debug info when debug flags are enabled. Error messages containing data that isn’t sensitive when it’s data you have access to anyways, but could be sensitive when the program has access to data that you don’t have access to. These situations expose methods that you can abuse certain programs to get flags.
  3. Sometimes, lazy programmers call out to other utilities instead of writing the functionality themselves. What happens if those other utilities are SUID root? What happens if you can influence where those other utilities are launched from (i.e., check out the PATH variable). This relies on lazy programming of the utilities that you are attacking, so you might want to pick up other targets before going down this path.

Further Reading

Some other useful resources: