Nebula exploit exercises walkthrough – level12

There is a backdoor process listening on port 50001.

My experience with Lua is minimal at best, but it’s pretty obvious that the hash() function calls a shell command, and allows for command injection.

To run getflag is very simple:

And if you want to pass the check for the hash for fun, it is also simple:

Nebula exploit exercises walkthrough – level11

The /home/flag11/flag11 binary processes standard input and executes a shell command.

There are two ways of completing this level, you may wish to do both πŸ™‚

Now it gets interesting. This is the first bit of code where it isn’t obvious what the intent is from a quick glance.

I think I have found three ways to get this to execute getflag, though one is just a variation of another.

The code reads from stdin, then checks for “Content-Length: “, reads a length integer, and then processes this.

There are a number of paths from this point. If the length is less than the buf length (1024), then fread is called. Then there is a bug.

This is what happens on this code path:

But later on:

From the man page of fread:

size_t fread(void *ptr, size_t size, size_t nmemb, FILE *stream);

The function fread() reads nmemb elements of data, each size bytes
long, from the stream pointed to by stream, storing them at the loca‐
tion given by ptr.

fread() and fwrite() return the number of items successfully read or
written (i.e., not the number of characters).

Whilet both read in the same data, the return values will be different. The first will return 1, the second will return the number of chars read.

This means the only way to get to process with the length less than 1024 is to set the length to 1. This restricts our options a fair bit.

We’ll try it out though:

As expected, the value we pass (E, arbitrary choice) gets “processed” to become D. system is then called, but because we can only provide a single character, we can’t null terminate the command, and we get some random values after.

We can see these values vary each time we run it:

One thing that does happen though is that, by chance, we end up with a null being in the right place:

This is pure luck. The rest of buffer is uninitialized and nulls are common in uninitialised memory.

If we now symbolic link D to /bin/getflag, and alter the path so it runs D when the null is in the right place:

Hmmph. Why is it not the flag account? I think this is a bug – the call to system isn’t preceded by setresuid/setresgid, so anything it runs will run as the real UID (level11) instead of the effective UID (flag11).

Co-incidentally, I had recently read of a technique to fill uninitialised memory. It’s virtually useless in the real world – using uninitalised memory is indicative of much bigger issues. It’s interesting though, so let’s try it here.

This technique uses an environment variable called LD_PRELOAD. This is commonly used to override library functions for debugging (or exploits!). When the linker starts up, it reads the entirity of LD_PRELOAD onto the stack and then doesn’t clean up afterwards. This means we can initialise the memory to something under out control:

i.e. fill the stack with one thousand /bin/getflags.

Then when we run flag11 with length of 1, it will almost certainly have this in the buffer already:

Again the same issue with suid/system, but I think it counts.

Now we need to come back to the length being 1024 or more. What happens here?

There is a really simple encryption function:

We can easily build the reverse of this in Python and output a string:

(note, that I terminated the command with newline (x0a) to start with, which was causing this to fail)

We can then pipe this into the executable, to run the command:



Whilst playing around with this level, I thougut there might be something I could do with the random path/filename that is used when the content length is 2014 or greater.

The filename is normally of the form:

As seen from strace. This is PID (process ID) with a “random” string.

We can gain control of this string, the filename, and stop it from being deleted. This uses LD_PRELOAD, but for it’s genuine use.

First, we must check that the executable is dynamically linked:


Now we need to create a c file to override the functions we want – random(), unlink() and getpid():

The we compile it into a library, set LD_PRELOAD, and then run the executable:

And now we have control of the filename, and it is preserved rather than deleted.

Not of any real use, but a handy technique.

Nebula exploit exercises walkthrough – level10

The setuid binary at /home/flag10/flag10 binary will upload any file given, as long as it meets the requirements of the access() system call.

I think I can already see the problem.

Firstly, we can see that the token file we need to read out is permissioned such that level10 cannot see it:

On line x above, we have the following:

From the man page of access:

access() checks whether the calling process can access the file path‐

The check is done using the calling process’s real UID and GID, rather
than the effective IDs as is done when actually attempting an operation
(e.g., open(2)) on the file.

So we check the file permissions using the real UID (level10), but then later on we do:

and open uses the effective UID, and as the executable has suid, this means flag10.

This is commonly called a time-of-use to time-of-check or TOCTOU bug (Wikipedia’s example is pretty much exactly the same issue)

If we can swap out the file between the time-of-check and the time-of-use, we should be able to send token.

First, let’s just check the program works as expected.

Setup a listening netcat on my host using:

And then run it on nebula with a file we have access to:

And we receive it at the other end, plus a little banner:

Ok – so how do we explout the race condition? The best way to swap the file about is to use symolic links again. How do we time that though? I’m fundamentally a lazy person, so let’s try and just swap out the files as quickly as we can and hope it works.

First, let’s setup a loop that flips a symbolic link from the real token to a fake one repeatedly:

The f switch on ln makes sure we overwrite the existing symbolic link. The &amp at the end puts the job into the background.

Then let’s setup the listening netcat to keep on listening rather than exit using the k switch.

And finally, let’s run flag10 repeatedly using another bash one-liner:

Go back to netcat and we have the token:

There we go – the password for flag10.

Nebula exploit exercises walkthrough – level08

World readable files strike again. Check what that user was up to, and use it to log into flag08 account.

A readable pcap file in the flag08 home directory. This is a network capture, so might have some interesting traffic.

Now… we can read this on the terminal using tcpdump:

Even when it is this prettied up, it’s still hard work – especially if it is a keyboard interactive process. People using the keyboard expect instant feedback – they press a key, they what to see the screen change. This means that there is a lot of back and forth. Compare this to, say, a request for a web page, which is machine generated and will fit neatly into packets.

So I want to get this file into Wireshark on my local machine. How can we do that? netcat!

(note that these instructions have OS X as the remote end – the command name and options syntax vary from OS to OS)

On the host machine, we do the following:

Listen on port 2001, and pipe any output to the file capture.pcap.

and on the client (Nebula machine) we do this:

Connect to port 2001 and pipe capture.pcap down the connection.

Now we have our file at the other end, it is an easy taste to run Wireshark and open the capture.Wireshark

There is a single connection between two given IPs here. The trace is still hard to follow though, so go to Analyze -> Follow TCP stream. This gives us a nice, coherent conversation:

We can see a login to another machine. We are just going to have to hope for some password re-use. The password bit looks like:

However, those . are not . – they are characters not represented by display characters. Switch the view to hex view and we can see:

Hex view

Hex view

x7f – DEL (well, backspace). That makes the password backd00Rmate

Nebula exploit exercises walkthrough – level07

The flag07 user was writing their very first perl program that allowed them to ping hosts to see if they were reachable from the web server.

The code of the CGI script is provided (and can be viewed in /home/flag07):



# check if Host set. if not, display normal page, etc


Immediately you can see this is not sanitising or validating the input parameter Host that it passes to a command – ping. We can therefore pass it another command for it to execute.

Let’s test the script out, from the command line to start with:

(I’ve stripped out HTML as I am lazy and can’t be bothered getting it to format correctly).

It just runs ping against localhost, as expected.

Run it without parameters, and we get the help:

And then let’s check we can inject a command:


The challenge now is that, for the first time, this script isn’t set to run suid. If I try running getflag, it isn’t going to work.

That thttpd.conf file in flag07’s home directory looks interesting. Could he be running a test web server?

Excellent – a web server on port 7007.

So, we need to:

  • Connect to the web server running on localhost at port 7007
  • Request a index.cgi
  • Pass a Host parameter with a command being careful to URL escape all of the special chars

wget is a simple utility present on nearly all Linux boxes that allows us to get a webpage.

We just need to escape the semi-colon to be %3B.

Check the content of the file and we have run getflag as a flag07.

Nebula exploit exercises walkthrough – level05

Check the flag05 home directory. You are looking for weak directory permissions

Let’s start looking in /home/flag05:

Compare to the home directory of level05:

So we have .ssh – the store of SSH keys for the user – and .backup. The .ssh directory is locked down so we can’t see it.

Let’s look in .backup:

A single backup .tgz. Let’s copy it out to our own home directory and unpack.

That’s the private (id_rsa) and public ( keys for flag05. They may well work on the local machine:

Simple. That’s why you should keep your private key private!

Nebula exploit exercises walkthrough – level04

This level requires you to read the token file, but the code restricts the files that can be read. Find a way to bypass it πŸ™‚

This program looks like it will read the file passed to it by the first argument. Let’s test that out:

Everything as expected then. The problem is that it explicitly forbids opening of files called token. How can we get round this?

Symbolic links to the rescue again!

Just create a symbolic link to a name that doesn’t match “token”.

So what is this long string? Seems sensible to try and login to the flag04 account with it:

Nebula exploit exercises walkthrough – level03


Check the home directory of flag03 and take note of the files there.

There is a crontab that is called every couple of minutes.

cron is a utility used to run tasks periodically, found in nearly every distro.

In /home/flag03, we have a script – – and a directory – writable.d.

Let’s take a look at

This is fairly simple – for each file in the writable.d directory, execute the scripts contained within, and then delete them. bash -x runs the script in a trace mode, to give you a bit more detail about when it is running. I think we can ignore ulimit -t 5 – it just limits the CPU time available to the shell, possibly to stop a malicious script consuming excess resources.

Note that the writable.d directory is world read/write – so we can just put a script in there:

Then wait a short while, assuming that the script is the one being run by cron…


Now – this is all well and good, but if we weren’t told that the script was run by cron, what could we do?

There is a root user in the Nebula VM, and using that I can do:

But I can’t do that as level03:

Also, I could use ps to see that the process runs, but that would presume that I knew it was cron’ed anyway.

So, not sure how I would go about finding cron jobs as an unprivileged user.

I’ve asked on the Unix Stack Exchange.

Nebula exploit exercises walkthrough – level02


There is a vulnerability in the below program that allows arbitrary programs to be executed, can you find it?

Another executable that calls system(). This time the command run is built up using an environment variable, USER.

Running the executable gives the expected result:

The executable is suid. Notice that although it calls system() and sets the setresgid()/setresuid() so that it runs as the owner of the file, the environment variable USER is still for the real UID, level02.

It’s really easy to change environment variables though.

This is a good reason to not trust environment variables for security purposes.


I didn’t fully understand why setresgid()/setresuid() had to be called for system() to run as the file owner. I built the same executable from source to experiment, set the owner, group and permissions as needed, but it didn’t work!

I spent a fair amount of time trying to figure this out, and it wasn’t until I did:

I was trying to run them out of /tmp/ and the whole directory doesn’t allow suid use…

Nebula exploit exercises walkthrough – level01


There is a vulnerability in the below program that allows arbitrary programs to be executed, can you find it?

The executable is located in the /home/flag01 directory. On running it, we get the expected output:

Importantly, if we check the permissions on the executable:

We can see that this file also has the suid bit set. The problem then is, how do we get this to run “getflag”?

The executable does nothing with command line parameters so we can’t pass anything in there. It does however call echo to output the text. echo is a built-in command to bash (i.e. not a discrete executable like ping would be), so we normally couldn’t override what it does.

However notice that the system call uses /user/bin/env before echo – where is this normally seen? At the start of scripts where we define the interpreter with a shebang.

The reason that /usr/bin/env is used is that scripts need a full path to the interpreter. python could be anywhere, and it is awkward to modify scripts to use a full path from system to system. /usr/bin/env searches the path for the command passed to it and runs it.

This means we can provide our own echo, modify the path so that this echo is called in preference to the built-in, and then we can run arbitrary commands.

The easiest way to provide our own echo that runs getflag is to just create a symbolic link.

Again – relatively simple. Symbolic links are useful tools for bypassing name and location checks!