Study Of Password Recovery Task’s Decision Mechanism On Heterogeneous Platform

With the development of computer technology, more and more people pay attention to the security of information. Password authentication mechanism is an important means to ensure the security of file. However, the password will be too complicated to the user’s memory burden, once forgotten password, may cause huge losses, while the encrypted file to the Security Department of the investigation and evidence collection work has brought inconvenience. Password recovery system is generated for this demand, which contains a large number of computing intensive module, the general computer’s computing power has been far from meeting the needs of the application. And heterogeneous architecture has become the trend of today’s high-performance computing platform, but there is a problem of high energy consumption. By adding the FPGA to the heterogeneous computing system, the system can not only reduce the system power consumption significantly, but also have a good effect on the computation intensive applications. This paper is based on principled sample machine of mimic computer and CPU+GPU+FPGA heterogeneous platform, the design covers MD5, SHA1, Office 2010, RAR and other encryption types of password recovery system, to achieve the improvement of the energy efficiency ratio.In order to improve the efficiency of the exhaustive crack, it is usually used in the dictionary file. However, there are some disadvantages of relying too much on dictionaries. First of all, dictionary files are composed of a series of regular or regular passwords, if the dictionary does not contain the password needed to hit will still be exhaustive crack, but this will reduce the efficiency of crack. Secondly, using the dictionary often have multiple, and the volume are often large, with the crack speed continues to accelerate, the core computing dictionary communication between the module and the storage module will encounter a problem, frequently reads the dictionary file will also reduce the efficiency of crack. Now, exhaustive mode and dictionary mode are relatively independent in the field of password recovery, and there are no study about their combination.In this paper, based on the original decision module of the password recovery system, a decision model based on filter Bloom is proposed. Using Bloom filter can succinctly represent a collection and efficiently query an elements are related to the set of features, and take Bloom filter query results as the password recovery tasks choose the dictionary or exhaustive attack method, which combines the two pattern organically. This method achieves crack efficiency maximization in CPU+GPU+FPGA heterogeneous platform. Finally, through the use of SHA1 encryption types of files to carry out experimental simulation.