A Technique Research In Dictionary Generation Of Passwords Based On Deep Learning

With the popularity of computers and the development of the Internet,security issues are also increasing.In response to the increasingly serious security problems,authentication technology is frequently used to distinguish between authorized and unauthorized users,thus ensuring system and data security.Password-based authentication is the simplest and most commonly used authentication method.Several studies have shown that passwords are non-uniform distribution.After analyzing a large number of user passwords,the technique of password dictionary generation mines the features of passwords,and then generates new password dictionary based on the features.The generated dictionary contains the most probable user passwords.In this paper,deep learning is applied to generating password dictionary for offline attack and online attack.Two dictionary generation methond is designed and implemented for non-specific user and specific user respectively.Non-specific user dictionary generation technique creates passwords with common feature extracted by recurrent neural networks.The experimental results,which shows higher coverage rate compared with the results of Markov and PCFG generation methods,proves that this technique can improve dictionary attack efficiently.Specific user dictionary generation technique uses recurrent neural networks to mine potential relationships between user passwords and user personal information,experiment results show that this technique can effectively guess user password structure and generates user passwords with personal information.Moreover,the experiment results show higher success rate compared with the results of Markov and PCFG generation methods and higher success rate compared with the once guess result of TarGuess method.