Abstract:

In order to improve the success rate of directional password guessing,this paper proposes a directional password guessing scheme based on the probabilistic context-free grammar (PCFG) combined with generative adversarial networks (GAN).First,this scheme matches the user’s real password with his personal information,and further divides the tags on the basis of the TarGuess-I model,and uses the divided tags to parse the real password.Second,the parsed passwords are input into the confrontation network,and an expanded rule set that follows the real password distribution is obtained after training.Finally,the guessed password set of the target user is generated according to the expanded rule set generated by training and the frequency table of the L (Letters),D (Digits),and S (Symbols) fields obtained from the user's personal information and the password parsing process.This paper adopts the method of demonstration based on the statistical results of the data to propose ideas and verification through experiments,and optimizes the innovative matching of the type-based personal identifiable information PII (Personal Identifiable Information):“numbers +letters” and “letters +special characters”.A series of studies on “numbers +special characters” (which should generally be “letters” and “numbers”) is carried out.Through guessing attack experiments on the railway 12306 data set containing users’ personal information,compared with other targeted password guessing schemes,this scheme has a higher success rate of password guessing.

Key words: password guessing, generative adversarial networks, natural language processing