Design And Implementation Of Enterprise Document Information Security Audit System Based On Fingerprint Detection

With the advent of the "Internet +" era,data has become the core asset of more and more organizations or enterprises.Documents are gradually transformed from paper version to electronic version,and the data transmitted and exchanged on the Internet are also increasing.The transformation of the way increases the risk and threat of data security.The development of modern enterprises has been promoted by means of information technology,but the security problems of sensitive data in enterprises have gradually emerged.For preventing the leakage of important documents,it is necessary to arrange effective detection methods for these sensitive documents in the enterprise.It is proposed a multi-file fingerprinting strategy,combined with advanced technologies such as artificial intelligence to design and implement the development requirements of the information security audit system.Firstly,by comparing the strengths and weaknesses of the several common classification models and analyzing the trait of data,the text classification module is designed.The enterprise text classification model based on Linears VC algorithm is built by iterating the information and adjusting and evaluating the model parameters for several times.This module is used to provide more accurate classification and prediction for the sorting of documents in the enterprise,and provide specification description for the detection of outgoing data.Secondly,in order to improve the security protection of documents by AES algorithm in enterprises,a data security module is designed.By combining two one-dimensional chaotic mappings into a compound chaotic map and combining with AES encryption algorithm,the AES encryption algorithm based on compound chaotic map is proposed.On this basis,the key space,diffusion and confusion of the traditional AES algorithm and the compound chaotic encryption algorithm are further analyzed.At the same time,the time performance of the algorithm proposed is tested and analyzed.Thirdly,in view of the insufficiency of traditional Simhash fingerprints for document detection,the recognition and detection module is designed,and three fingerprint algorithms of Pb S,Kb S,and So P are proposed according to the different feature extraction methods.By conducting simulation experiments on sensitive documents with different degrees of deletion,addition,mixing and other operations,the Hamming distance distribution between documents with different degrees of modification and original documents was explored,which providing the basis for sensitive threshold range of multiple document fingerprint strategy.Finally,the function and performance of the sensitive information audit system is tested.The efficiency of fingerprint detection strategies in the system is tested by the experiment of outgoing documents to simulate the real situation,and the classification effect of enterprise text is tested by the trained model.The accuracy,recall and F1 value of the three indexes were analyzed to verify the practical efficiency of sensitive document detection and the accuracy of text classification prediction of the system.