Research On Distributed Cryptographic Analysis Platform Based On Heterogeneous Computing

With the development of the information society,information technology gradually integrates into people's daily lives,but malicious cyber attacks often steal user's privacy and pose a threat to user's security.In order to protect user's network security,various encryption technologies have emerged.The diversity and personalization of encryption algorithms have led to a great increase in difficulty in cracking passwords.In cryptanalysis and computer security,the cracking efficiency of most software is usually low,and it takes a long time to waste computing resources.However,the advent of GPUs and supercomputers has greatly improved the computing power of computers.Modern GPUs can be used in highly parallel structures to make them more efficient in processing large blocks of data.Heterogeneous computing is also gradually applied to scientific research and practical projects to improve computational efficiency.This paper uses the current mainstream machine learning model:LSTM and TextCNN learn the dependencies of the sequence of cipher data sets.Using the adaptive moment estimation random gradient descent method,the data set is trained,and the problem of disappearance of gradients due to long-distance dependence is solved.The random generation algorithm and the global generation algorithm were designed to generate the password dictionary from the two aspects:the efficiency of dictionary generation and the coverage of the dictionary.The system can also generate personalized dictionaries based on user-defined custom password data sets,making the dictionaries more targeted.This paper also designs a distributed cryptanalysis platform based on heterogeneous computing platform.At present,the cryptographic analysis platform can support brute-force cracking and dictionary cracking.After the transformation of the open-source password cracking tool Hashcat,it was transplanted to the Changsha Tianhe-1 Supercomputer Platform,the Guangzhou Tianhe-2 Supercomputer Platform,and the laboratory's self-built cluster.Due to the low coupling of password cracking,the tasks can be parallelized when the password is cracked to improve the computational efficiency.Using distributed computing technology,the password task to be cracked is divided into independent sub-tasks and each node on the cluster independently performs calculations.The MPI message communication mechanism is used to communicate between nodes,which improves the cooperation efficiency between nodes.Combine this cryptanalysis platform with the generated intelligent dictionary to improve the efficiency of dictionary cracking.Efficient password cracking technology can help users recover forgotten passwords.The generated high-frequency password dictionary can also help users to avoid such passwords to improve the security of the user setting passwords.From the perspective of national security,the high efficiency of this system can quickly obtain effective information for the security sector to severely crack down on acts that endanger national unity and national security.