asleap - recovers weak LEAP password. Pronounced "asleep". Copyright(c) 2004-2008, Joshua Wright $Id: README,v 1.10 2007/05/10 19:29:06 jwright Exp $ -------------------------------------------------------------------------------- UPDATE - 2008-5-28 Minor update to fix a problem with IFNAMSIZ errors in build on some platforms. UDPATE - 2007-7-13 I updated asleap to include the "-C" and "-W" options, where you can specify the challenge value (in colon-delimited bytes) and the response (ditto) on the command-line. This makes asleap more of a generic MS-CHAPv2 dictionary attack tool, which is fine by me. Why did I do this? Because Jay "F***ing" Beale asked me to. Nicely. UPDATE - 2007-5-10 I spent some time updating asleap and came to the following conclusions: + My code is much better now than it was in 2003/2004 + I was smoking crack when I wrote this the first time I removed a lot of functionality that I didnt't think was necessary any more in this version, specifically: + No more support for Airopeek NX files; if you want to read from a .apc file, install Wireshark, and run "tshark -r input.apc -w output.dump" to convert to libpcap format. + No more integration with Airjack to deauth users. If you still want to deauth users to get their LEAP credentials without waiting, you can use a tool such as file2air, MDK2 or aireplay-ng. + I stopped caring about Windows a long time ago. If you need this, and you only run Windows, try using a bootable Linux ISO like Backtrack (www.remote-exploit.org) and VMWare. On the brighter side: + The code has 90% less teh suck + Fixed an awful bug where passwords > 64 characters caused genkeys to segfault. + Added code to handle QoS data frames + Can handle radiotap-formatted capture files now NB: The code to handle QoS data frames isn't tested. Why you ask? Because Zero_Chaos didn't send me the packet capture I asked for. pfppbbbbbbb What follows is the original README. Please contact me with any questions, comments or QoS LEAP transactions. jwright@hasborg.com -- INTRO This tool is released as a proof-of-concept to demonstrate weaknesses in the LEAP and PPTP protocols. LEAP is the Lightweight Extensible Authentication Protocol, intellectual property of Cisco Systems, Inc. LEAP is a security mechanism available only on Cisco access points to perform authentication of end-users and access points. LEAP is written as a standard EAP-type, but is not compliant with the 802.1X specification since the access point modifies packets in transit, instead of simply passing them to a authentication server (e.g. RADIUS). PPTP is a Microsoft invention for deploying virual private networks (VPN). PPTP uses a tunneling method to transfer PPP frames over an insecure network such as a wireless LAN. RFC 2637 documents the operation and functionality of the PPTP protocol. BACKGROUND LEAP utilizes a modified MS-CHAPv2 challenge/response in order to authenticate users on a wireless network. The MS-CHAPv2 authentication method has been clearly identified as a weak method of authentication for several reasons. A standard LEAP exchange is as follows: 1. STA requests authentication with 802.1X start message 2. AP issues a random 8-byte challenge 3. STA encrypts the 8-byte challenge 3 times, using the NT hash of their password as seed material. The STA then joins the 3 DES outputs as a single 24-byte response. 4. AP issues a success or failure message. 5. STA issues a 8-byte challenge. 6. AP responds with a 24-byte response. 7. STA is able to send data to the distribution system. PPTP networks use a straight implementation of MS-CHAPv2 for authentication, which is slightly more complex than the LEAP example: 1. STA initiates PPP configuration negotiation options 2. PPTP server (SERVER) issues a random 16-byte challenge 3. STA generates a random 16-byte STA challenge 4. STA uses SHA1 to hash the SERVER 16-byte challenge, the ST challenge and the STA's username. The first 8-bytes become the 8-byte random challenge (step 2 in LEAP). 5. STA encrypts the 8-byte challenge 3 times using the NT hash of their password as see material. The three 8-byte outputs are contcatenated to generate a single 24-byte response. 6. STA transmits the 24-byte response, the 16-byte STA challenge and the STA username to the SERVER. 7. SERVER decrypts the 24-byte response with the STA's stored NT password hash. If the decrypted response does not match the SERVER challenge, the SERVER issues a PPP CHAP Failure message. 8. If the response does match the SERVER challenge, the SERVER hashes the STA password hash to create a password-hash-hash. 9. The SERVER combines the password-hash-hash, the literal "Magic server to client signing constant", and the 24-byte response and calculates a SHA1 hash. 10.The SERVER calculates the 8-byte challenge using the same procedure in (4). 11.Combining the output from (9) and (10) with the literal constant "Pad to make it do more than one iteration", the SERVER calculates another SHA1 hash. 12.The output of (11) is converted to upper-case ASCII-hexadecimal representation and transmitted back to the STA. 13.The STA uses steps (8-12) to authenticate* the SERVER. * If anyone can explain to me why the MS-CHAPv2 algorithm has all this seemingly unnecessary complexity, please contact me. Unfortunately, the MS-CHAPv2 challenge/response suffers from several notable flaws: + No salt in used in conjunction with the NT hash - Permits pre-computed dictionary attacks + Weak DES key selection for challenge/response - Permits recovery of 2 bytes of the NT hash + Username sent in clear-text With these flaws present, we are able to deduce enough information to mount a dictionary attack against a user's password after capturing their successful login to a LEAP or PPTP network. Since we are attacking a wireless network exchange, this is a trivial task with a wireless sniffer. OUR ATTACK Since there is no salt in the NT hash, we can precalculate an indexed list of passwords and their associated NT hashes. This is only marginally useful to us, since we would normally need to run three rounds of DES to encrypt the random challenge for each word in our dictionary. Not an unreasonable task, but certainly time-consuming on a single mobile device like a laptop or a PDA. We can significantly reduce the amount of time needed to launch an attack against the LEAP or PPTP challenge/response by leveraging a second flaw in the MS-CHAPv2 protocol. When a STA receives a challenge, they encrypt the challenge three times, using portions of their password hash as DES seeds. The DES algorithm requires a 7-byte seed value in this algorithm, so the STA splits their 16-byte NT hash into three portions: seed1 = HB1 .. HB7 seed2 = HB8 .. HB14 seed3 = HB15, HB16, 0x00, 0x00, 0x00, 0x00, 0x00 Where HBx is the bytes of the NT hash, and 0x00 is a standard NULL byte. The flaw with this method is that the third DES output is cryptographically weak, leaving only 2^16 possible permutations. This can be calculated in a matter of seconds, even on older Pentium hardware. By calculating all 2^16 possibilities of the last 8 bytes of the response value, we can derive the last 2 bytes of the NT hash (HB15, HB16). With the last two bytes of the NT hash, we can significantly reduce the number of potential matches in our dictionary file. We can also utilize this information to perform highly optimized indexed lookups on our dictionary file, using the known 2 bytes of the users hash as a keying mechanism. Asleap will examine traffic to identify the LEAP or PPTP challenge/response process and will extract the username, challenge and response information. This can be done in a passive operation with a 802.11 card in RFMON mode, or through an active attack, where asleap impersonates the legitimate access point, and forces the victim to deauthenticate from the network. From the STA response and the challenge, asleap will calculate the last two bytes of the matching NT hash, and display that value. After calculating the last two bytes of the NT hash, asleap will search the genkeys output file of dict+hashes for matching hash values. With the matching hash value, asleap will use it to seed the DES operation and encrypt the challenge, comparing the output to the captured response from the STA. If the calculated 24-byte response value matches the captured 24-byte response, asleap will display the associated clear-text password. If the password for the LEAP or PPTP user is not present in your dictionary file, asleap will not be able to report the user's password. The quality of your dictionary file will have a lot to do with the success of recovering LEAP/PPTP user passwords. THE TOOLS You will need a dictionary file, one word per line, to use with asleap. Please don't ask me for my dictionary file - be creative and put your own together. You can use the morewords.pl and strip-nonascii.pl scripts in the scripts/ directory if desired. + STEP 1. Generate the database and index files to use with asleap lookups by running them through genkeys. Very straightforward: $ ./genkeys genkeys 2.0 - generates lookup file for asleap. genkeys: Must supply -r -f and -n Usage: genkeys [options] -r Input dictionary file, one word per line -f Output pass+hash filename -n Output index filename -h Last 2 hash bytes to filter with (optional) $ I typically use .dat and .idx extensions on the database ("outfile") and index file, just to keep it straight in my head. + STEP 2. Decide how you are going to collect LEAP credentials for use with asleap. This could be through a live interface in RFMON mode, or through a previous capture file in libpcap format. Take a look at the command line options for asleap: $ ./asleap asleap 2.0 - actively recover LEAP/PPTP passwords. asleap: Must supply an interface with -i, or a stored file with -r Usage: asleap [options] -r Read from a libpcap file -i Interface to capture on -f Dictionary file with NT hashes -n Index file for NT hashes -s Skip the check to make sure authentication was successful -h Output this help information and exit -v Print verbose information (more -v for more verbosity) -V Print program version and exit -W ASCII dictionary file (special purpose) $ BUGS I'm certain. Let me know about them. While not a bug - a word on channel hopping. Use channel hopping only to look for networks that are running LEAP. If you are always hopping to a different channel, you are likely going to miss valid targets. THE FIX LEAP is inherently weak as it relies on MS-CHAPv2 for authentication of clients on the wireless network. Cisco seems to acknowledge this fact in their support of the Protected Extensible Authentication Protocol (PEAP) that uses MS-CHAPv2 authentication within a stronger TLS encryption channel. Users relying on LEAP for WLAN access control should consider switching to PEAP or TTLS, both open EAP types submitted to the IETF for standardization. As a short-term measure, LEAP users should immediately audit the passwords of their users to identify weak passwords. Weak passwords should be expired and replaced with passwords at least 8 characters in length that consist of numbers, letters and special characters. This is only a stop-gap measure, since an attacker with sufficient disk space could create an exhaustive list of all printable characters to use for their dictionary attacks. Since PPTP uses the same MS-CHAPv2 algorithm, it suffers from the same weaknesses as LEAP. Microsoft recommends users select very strong passwords, or switch to a more secure authentication mechanism such as IPSec or L2TP. QUESTIONS, COMMENTS, CONCERNS Please contact jwright@hasborg.com with any questions, comments on concerns. My PGP key is located at http://802.11ninja.net/~jwright/pgpkey.html