This weekend I participated in the GPN CTF, organised by the CTF team of the Karlsruhe Institute of Technology. I focused on the challenges tagged "Introduction", solving all except the Pwning one. For my readers unfamiliar with CTFs, I recently wrote this post about their role in my career so far. The flags are listed in the order I got them.
It's customary for CTF organisers to include at least one trivially easy flag to alleviate the participants worries that subsequent flags are rejected due to the system not working. As hinted at in the title, the flag was present in full on the site listing the rules for the event.
This challenge gave a Dockerfile and some PHP code.
The Dockerfile told me there was a file called flag.txt
under the root directory, making the objective pretty obvious.
What was less obvious was how to talk to the server and how the hell this PHP worked.
Talking to the server turned out to be not too complicated after looking up the appropriate curl options to arrive at something like this:
curl -A "friendlyHuman" --data 'data={"user": "adminπ€ ", "password": "something"}
The PHP wanted me to meet the strange condition that my supplied password be equal to the result of applying the function securePassword
to it.
function securePassword($user_secret){
if ($user_secret < 10000){
die("nope don't cheat");
}
$o = (integer) (substr(hexdec(md5(strval($user_secret))),0,7)*123981337);
return $user_secret * $o ;
}
So my first intuition was that 0*x=0
and surely that must be useful to find this flag.
I ended up spending about two hours first doing some napkin math of how long it would take my iGPU to find an md5 hash that starts with seven zeroes, followed by some Hashcat related research and ultimately finding the hashgame.
Somewhere along the way I heard about PHP type juggling and that in PHP, due to it, md5('240610708')
equals md5('QNKCDZO')
.
After a short break I fired up another instance to get the communication rolling again.
In my infinite wisdom, I supply the password "11e777777" and am surprised to be greeted as admin.
From here I just had to add "cat /flag.txt" to the "command" field of my json POST data.
This challenge gave a binary that runs a hash function as well as the hashed flag.
At first I ran the binary through strings
and xxd
, and with strace
and ltrace
to no avail.
I then tested various inputs, quickly noticing that the hashing function is linear. What do I mean by that? Well, a character in the input string maps to a fixed number of digits in the output string, usually two. The early characters influence later characters, but not the other way around. If there was no influence at all, we could quickly build a substitution table, but this is still fine.
To get the flag I wrote a python script that builds up the flag, checking each character for matching hash output as it appends it. The code speaks for itself:
import subprocess
target_hash = "4717591a4e08732410215579264e7e0956320367384171045b28187402316e1a7243300f501946325a6a1f7810643b0a7e21566257083c63043404603f5763563e43"
flag = "GPNCTF{"
local_hash = ""
while not local_hash == target_hash:
for x in range(32, 127):
local_flag = flag + chr(x)
local_hash = subprocess.check_output(["./hasher", local_flag]).decode()[:-1]
if(all(x == y for x,y in zip(local_hash, target_hash))):
flag += chr(x)
break
print(flag)
This challenge gave a chal.py
and a file with the encrypted flag.
chal.py
contained an implementation of the XOR cipher, with the notable weakness being the key length of 5.
The XOR cipher is uniquely symmetric in that it's not just a symmetric cipher, but one where the encryption function is identical to the decryption function.
Since the flag format was known to match GPNCTF{.*}
, we had a known plaintext.
With this, and after fighting pythons type system for a little, I was able to derive the key and then use that to decrypt the flag.
Check out the code:
flag_enc = "d24fe00395d364e12ea4ca4b9f2da4ca6f9a24b2ca729a399efb2cd873b3ca7d9d1fb3a66a9b73a5b43e8f3d"
kpt = "GPNCT"
# convert to bytes
b_flag_enc = bytes.fromhex(flag_enc)
b_kpt = kpt.encode("utf8")
# derive the key
key = []
for i in range(len(b_kpt)):
key += [b_flag_enc[i]^b_kpt[i]]
key = bytes(key)
# get flag
flag = []
for i in range(len(b_flag_enc)):
flag += [b_flag_enc[i]^key[i%len(key)]]
print(bytes(flag).decode())
With this challenge you're tasked with uploading a file to a website.
The website then executes the uploaded file, with a catflag
binary available in the root directory of the server.
I'm not sure I could have done this challenge without hints.
First order of business was to write a hello world program in C, which went well.
Then, after a little digging, I thought execl
would do the trick, calling /catflag
from within my binary.
While it worked locally to the extend testable, it did not yield a flag.
Checking the chat logs, I came across the recommendation to use patchelf
.
So I thought to add /catflag
as a dependency via patchelf --add-needed "/catflag" a.out
, but that wasn't enough.
What was actually necessary was to use --set-interpreter
instead of --add-needed
.
The idea for that came from this post, but I'm sure the patchelf hint would have been sufficient with a little more perseverance.