OneRuleToFindThem: Efficient Automated Generation of Password Cracking Rules

Password cracking tools such as Hashcat support the use of rules that transform a dictionary of words, such as common English words and previously-cracked passwords, into new candidate guesses for hashed passwords. Rules are necessary to achieve high cracking ratios, but they are difficult and time-consuming to build by hand. We have developed an algorithm and implementation that automatically finds successful rules via the combinatorial generation of rules and empirical observation of how often each generated rule transforms a dictionary word to a target password. Our algorithm includes numerous performance and logical optimiza- tions to avoid the numerous pitfalls that would occur if a naïve brute- force technique were used. In this paper, we explain our algorithm in detail and experimentally compare the performance of its outputs to existing rule sets constructed via various approaches ranging from fully- manual to fully-automated like our own. We show that our approach achieves comparable cracking perfor- mance to other top rule sets while also generating rules that do not exist in other rule sets. This makes cracking attempts using our rules mostly complementary to cracking attempts with other rule sets. We demon- strate that we can achieve top performance by combining our generated rules with other rule sets.

Recommended citation: J. Eckroth, L. Hough, H. ElAarag. OneRuleToFindThem: Efficient Automated Generation of Password Cracking Rules. CCSC Eastern 2023.
Download Paper