Gone in 120 seconds: cracking Wi-Fi security
WEP is dead
What can affect the speed of your attack?
Erik Tews: There are some keys we call strong keys. A key is a strong key if it has at least one strong keybyte. A strong keybyte is a keybyte which fulfills a special equation or condition. (Equation (10) in section 6.2 in our paper)
If a key has just 1-3 or perhaps 1-4 of these strong keybytes, our attack will still work, but perhaps take some more packets. The probability that a randomly chosen key has more strong keybytes is below one per cent.
Even if a key has the maximum of 12 strong keybytes, our attack can be modified so that it will still work, just need some more cpu-time or packets. This is currently not implemented in our tool, but we know how to fix that and we are going to implement it soon. With our modification, we will perhaps need three to five minutes with an optimal packet rate for a key with 12 strong bytes (this is a guess, hasn't been exactly tested yet).
What about the keys with a bigger size than 104 bit?
Erik Tews: There are some vendors which implemented a 232/256 bit WEP. I think these keysizes are very uncommon. Currently, only 40/64 and 104/128 bit keys have been implemented.
There are currently some other attacks, which allow us to recover more than the first 16 bytes of the keystream. Combining our attack with these attacks would even allow us to break WEP512. This has not yet been implemented, but could be added in future.
How does your attack performance scale with increasing WEP key size?
Erik Tews: We did only benchmark the 104 bit version of WEP. If just a 40 bit key is used, we know the attack is faster, but we didn't do exact benchmarks. Perhaps it can be done in 30 seconds if the packet rate is high.
Do 256 bits stop you from using just ARP packets to succeed?
Erik Tews: For an ARP-Response, the first 16 bytes are constant. What follows are IP and MAC-Adresses. These values are not globally fixed, but if the same request is sent again and again, these values will be always the same because the response is the same again and again.
There is another attack called chopchop which should be able to find out what these unknown values are. On the other hand, these values could perhaps be guessed too.
Using such a technique, it should be possible to attack WEP256 too. This is currently not implemented in aircrack-ptw, but could be added easily.