PASSWORD CRACKING BASED ON LEARNED PATTERNS FROM DISCLOSED PASSWORDS

Password-based authentication systems are still the most commonly used mechanism for protecting sensitive information despite being vulnerable to dictionary based attacks. To guard against such attacks, many organizations enforce complicated password-creation rules and require that passwords include numeric and special characters. This study demonstrates that as long as passwords are not difficult to remember, they remain vulnerable to “smart dictionary” attacks. In this study, a password analysis platform is developed to formally analyze commonly used passwords and identify frequently used password patterns and their associated probabilities. Based upon these patterns, we establish a model consisting of a Training set, a Dictionary set and a Testing set (TDT model) to generate probabilistic passwords sorted in decreasing order. The model can be used to dramatically reduce the size of the password space to be searched. Simulation results show that the number of passwords cracked using the TDT model is 1.43 and 2.5 times higher compared with the John-the-Ripper attack and Brute-force attack, respectively. We also design a hybrid password cracking system combining different attacks to verify the effectiveness of the proposed method. After applying the TDT model, the number of passwords cracked increased by up to 273%.