Research And Implementation Of The Auxiliary Secret-level Setting System Based On Deep Learning

At present,the secret level setting work of various government agencies in our country is mainly carried out manually by the responsible person,which may easily lead to problems such as irregular secret level setting procedures,low efficiency of secret level setting,and "slapping the brain to determine secret level".These problems are hidden dangers in confidential work and may lead to cases of loss of confidentiality.The use of computer technology to solve the difficulties and pain points in the secret level setting work is the research goal.A set of computer-aided secret level classification system is designed and implemented,which has the characteristics of auxiliaryness,accuracy,timeliness and flexibility,and can effectively assist the secret level setting workers to carry out the level setting work.An in-depth investigation is made on the current research status in the field of auxiliary secret level classification system.In view of the bottlenecks and difficulties in this field,using natural language processing technology and deep learning technology,two methods of auxiliary secret level classification are proposed.The main contents are as follows:(1)An auxiliary secret level setting method based on item matching is proposed.First of all,this method conducts content research and sorting out the "Contents List of Confidential Items",and organizes it into a tree structure,transforming the secret level setting process into a path selection process in a tree,thereby standardizing the secret level setting process.Secondly,this method uses the improved Text Rank method to extract secret feature word information in existing classified documents,and uses semantic matching method based on BERT model to extract semantic information.Finally,combining the features of both feature words and semantics,weightedly calculate the matching degree between the document to be classified and the confidential items,and recommend the confidential items for the secret level setting workers.Experimental results show that this method can accurately complete the item matching task.(2)An auxiliary secret level setting method based on Text CGA network is proposed.This method is suitable for documents whose confidential item matching results are not ideal.First,the method regards the secret level setting task as a text classification task,and applies deep learning techniques to the auxiliary secret level setting ask.Secondly,the CGA module for extracting text features is proposed,which solves the long-term dependence problem in regular RNN network to a certain extent,and enhances the network’s ability to extract local text features.Finally,the Text CGA text classification network was built using multiple CGA modules configured with different parameters,further enhancing the feature extraction capability of the network.The results of the performance experiment show that the network has good text classification capabilities;the results of the simulation secret level setting experiment show that the Text CGA network can complete the auxiliary secret level setting task well.auxiliarynessThe above two methods are integrated,and the auxiliary secret level setting system is implemented according to the design principles of auxiliaryness,accuracy,timeliness and flexibility,and the system is tested and analyzed through experiments.The test results show that the auxiliary secret level setting system standardizes the level setting process of the responsible personnel,can give accurate opinions on level setting,and can well complete the auxiliary secret level setting task.