| The rapid development of the Internet has led to a rapid growth in the number of Internet users worldwide.In order to obtain personalized data for different users,most platforms require users to use account and password to obtain their application resources or services on this platform,which makes password the main authentication method in today’s social networks.The traditional password dictionary construction algorithms are mostly based on empirical knowledge and artificially set rule sets,and have achieved good results in the application of password guessing.However,with the development of the Internet,the complexity of user password setting rules increases,and the effectiveness of password dictionary construction based on artificially set rules gradually decreases.With the development of the Internet,many plaintext password data sets have been leaked with various security events.The leaked user password also reflects some information about the user setting password.With these data sets,more and more researchers are interested in learning to disclose passwords through algorithms,so as to infer users’ habit of constructing passwords,and then generate an effective password dictionary.They have carried out a variety of research in password attacks,password generation,user password habits and so on.This paper believes that humans have certain selectivity in setting passwords,so the password string is regarded as a text sequence.Based on the existing long-term and short-term memory artificial neural network(LSTM)algorithm,and the random gradient descent method is selected to learn and train the leaked password data set of rockyou company.The improvement of modern GPU computing power can more efficiently process a large amount of data,and GPU is used to accelerate the model training process.This paper compares the coverage of the cipher dictionary generated by the long-term and short-term memory artificial neural network algorithm with the existing Markov chain cipher generation model and the cipher dictionary generated by the probabilistic context free method.Select the trained model to generate the password dictionary according to certain methods to improve the performance of the traditional password dictionary attack and increase the coverage of the password dictionary for the test data set;At the same time,this paper improves the traditional method of determining the strength of the password according to the length of the user’s password and the number of characters contained,and proposes a password strength meter based on LSTM algorithm model to help users reduce the contradiction between password strength and memory difficulty,so that users can increase password strength and improve user experience while facilitating memory.Finally,some suggestions for future password setting to enhance the security of user passwords and make it difficult to crack. |
