Research On Semantic Role Mining Based On Parallelized TF-IDF Algorithm In Large Data Environment

In the field of computer and information security. Access control technology has been a very important hot field. The system guarantees system security by limiting the user's access to resources within the system. RBAC (Role Based Access Contral) technology emerged from 1996 and gradually formed specifications. And now has become the standard of access control standards. But with the development of computing technology. System resources and system users started explosive growth. The need for system administrator configuration is also increasing. The pure artificial division RBAC access is becoming more and more expensive. In order to solve the above problems.Implement semi-automatic or automated role access control Role mining technology has gradually become the hot topic in this field.But traditional role mining is based on user permission and 0/1 mapping matrix.There are still many problems such as mining results without semantic information.And mining 0/1 matrix is not suitable for large amounts of data, easy to generate memory overflow and poor noise immunity problems. In view of the above problems, considering the influence of the role mining results and the actual production environment. In view of the above problems,considering the influence of the role mining results and the actual production environment.This paper proposes the method of changing the problem space,using topic mining method to build RBAC system.In this paper,we design a topic oriented model for role mining in real system .That is to use enterprise production log data to do the role mining and change the problem space of traditional role mining to identify roles. This paper constructs a user permission data set based on topic mining model by reconstructing the role mining data set. And user rights text data is mined using the classic topic mining algorithm TF-IDF. which is identified as the user's topic through the topic of each user file. Constitutes a user role authority result set.Then, in view of the current large data problem. This paper implements the distributed parallelization of the TF-IDF algorithm through the MapReduce computing framework of Hadoop platform. And proposes the distributed TF-IDF role mining. Finally, this paper verifies the performance of the algorithm in time performance and accuracy of mining results by comparing experiments. Experimental results show that the algorithm has better time performance and higher noise immunity, and mining has the advantages of semantics.