CSAW CTF 2012: Web 400 writeup
Written by Pierre Bourdon
Web 400 was an interesting challenge involving web exploits as well as crypto. We had access to a web application which allowed sending messages from a user to another. The twist is that all of these messages were encrypted using an unknown algorithm. When sending a message the user provides a key which is used to encrypt the message.
After analyzing the algorithm a bit (same key and message, trying different key sizes and block sizes, checking if every block is encrypted the same, etc.) we found out that it was some kind of ECB XOR using the key + a constant 64 bits value. This was only true for the first few blocks though: after that another key or another constant value was used. As we'll soon see, this does not matter a lot.
We were able to confirm that this message system is vulnerable to XSS attacks by sending some strings that give HTML tags when encrypted. We just need to encode a cookie stealer and send it to the admin user to gain access to his account.
Now that we know this algorithm uses XOR as its main operation, we can use a very interesting property of this binary operator:$$Plain \oplus Key = Cipher \Leftrightarrow Plain \oplus Cipher = Key$$
If we send a block using a plaintext
P1and it gives us
C1, we can use that property to deduce what we should send to have
C2be what we want:$$P2 = C2 \oplus Key \Rightarrow P2 = C2 \oplus (P1 \oplus C1)$$
It turns out we can't use that for a whole message because the key seems to depend on the previous blocks plaintexts. We had to construct the message block per block using that technic. When encrypted, our message is:
We sent that to the admin and got his session ID transmitted to our server. Using that we were able to login to his account and find some encrypted messages (and their associated key). The first message had a plaintext key when decrypted gave us another encryption key, which we used to decrypt a second message, giving us the final key we had to submit on the CTF website.blog comments powered by Disqus