Testing Password Encryption Strength
Testing the Password Encryption Strength of NT LAN Manager and LAN Manager Hash
Security risks associated with weak user-created passwords are well documented. In 2009, for example, cyber security provider Imperva analyzed more than 32 million passwords that were released in a 2009 data breach. More than 50% of the passwords involved poor user password choices, and 30% of the passwords contained 6 or fewer characters. User habits in poor password construction improve the chances of successful password determination by hackers who use password-guessing software.
Kivu recently participated in an experiment to evaluate the password encryption strength of two Windows Operating System authentication protocols.
LAN Manager (LM) hash, employs a multi-step algorithm to transform a user password into a calculated string value that obfuscates a password’s identity. The resulting LM hash is stored rather than the original password. First, a user’s password is converted to all uppercase letters. Next, the uppercase password is set to a 14-byte length. For passwords greater than 14 bytes, the password is truncated after the 14th byte. Passwords less than 14 bytes are null-padded to reach 14 bytes. The 14-byte password is split into two 7-byte segments, and a null is added at the beginning of each 7-byte half. Each half is used as a key to DES-encrypt the ASCII string “KGS!@#$%”. Both output values are concatenated to create a 16-byte value LM hash.
Microsoft’s second encryption method, NT LAN Manager (NTLM), is an improved algorithm for securing a password’s identity. Beginning with Microsoft Windows NT 3.1, NTLM was introduced to improve security. NTLM passwords differ from LM passwords in that NTLM employs the Unicode character set with the ability to differentiate upper and lowercase letters and permits passwords up to 128 characters in length.
Kivu’s experimental design to compare the relative strength of these two encryption methods employed Cain and Abel password-guessing software and different Windows passwords of increasing complexity.
How Cain and Abel Works
Cain and Abel provide the ability to execute password-guessing schemes using dictionary attacks and brute-force attacks. Both dictionary attacks and brute-force attacks employ guess-based methodologies to identify the plain text password associated with a specific hash-encrypted password.
Dictionary attacks use a pre-defined list of search terms or phrases as the basis for guessing. Each search term is transformed into a hash string value using a specific hash algorithm, such as the LM hash protocol. The resulting hash value is compared to a hash value of interest, and if the hash values match, the plain text password is identified.
Brute force attacks attempt every combination of defined search criteria to identify the plain text password associated with a hashed password. Search criteria settings include the use of character sets, such as ASCII and the number of characters in a password.
Conclusions
Kivu’s experimental results yielded significant insights concerning the strengths of the NTLM password algorithm, which is Microsoft’s replacement to LM.
None of the NTLM-transformed passwords we used were quickly resolved through a brute force attack. Our experiment suggested that the passwords established for the test user accounts would take more than 4 years to determine. While our brute-force attacks were limited to less than 3 minutes, NTLM hash protocols were identified as having substantial lead times to identify plain text equivalents of NTLM-hashed equivalents.
LM’s password hashing approach to obfuscating plain text passwords, however, was limited in its success. As observed with three test user passwords, brute force password guessing resulted in a partial identification of LM passwords, due to LM’s sub-division of the password string during the hashing algorithmic process.
Our results indicated that both NTLM and LM passwords are susceptible to compromise in a well-designed and broad dictionary attack. Overall, NTLM hashed password equivalents may be stronger than LM in simple dictionary attacks. In substantial dictionary attacks, however, it may be more likely to identify an NTLM password due to the ability to match a calculated hash value from a dictionary. While dictionary-based attacks may be limited in their combinations of matches, larger dictionaries provide more opportunities for a match.
Kivu (www.kivuconsulting.com) is a nationwide technology firm specializing in forensic response to data breaches and proactive IT security compliance. Headquartered in San Francisco with offices in Los Angeles, New York, Washington DC, and Vancouver, Kivu handles assignments throughout the US and Canada and is a pre-approved cyber forensics vendor for leading North American insurance carriers. Author, Megan Bell, directs data analysis projects and cyber security investigations at Kivu.