Research On Automatic Scoring Of Evaluation Report Of Information System Hierarchical Protection

With the rapid rise of the Internet,our country's network infrastructure is accelerating construction.Informatization has brought a golden opportunity for our country,but it also brings huge security challenges.In order to ensure the long-term,safe and orderly development of our country's network construction,the government has proposed a hierarchical protection system,which is regarded as a national standard for hierarchical and classified protection of information security systems.Hierarchical protection evaluation is the core part of hierarchical protection work.Improving the efficiency of evaluation work has important practical significance for promoting the development of network security.This paper takes the automatic scoring of grade protection evaluation report as the research object.Based on the analysis of automatic scoring research at home and abroad,it explores the method of realizing automatic scoring on the grade protection evaluation data.Based on the analysis of automatic scoring research at home and abroad,this paper explores the method of realizing automatic scoring on the grade protection evaluation data.The feasibility study is mainly conducted from two perspectives of text similarity and text classification.The effect of the experiment on text similarity is not very satisfactory,so the experiment finally chooses the text classification method for the study of automatic scoring of the evaluation report.The category to which the text belongs corresponds to the corresponding score,so as to realize the automatic scoring of the text.The main tasks of the thesis are as follows:(1)Construct a data set of evaluation records.The graded protection evaluation data is non-public data.The data set needs to be constructed by myself.The data used in the article is compiled from the graded protection evaluation report provided by the laboratory.It is a true record of the actual security status of the information system,and sensitive information is desensitized.Used as training corpus after processing.(2)Automatic scoring based on support vector machine,using bag-of-words model to represent text,and TF-IDF algorithm for text feature extraction.In order to improve the effect of the model,an improved TIF-NG method based on Ngram and an improved TIF-NG-IG feature extraction method based on information gain are also proposed.In the classification model,traditional machine learning models such as naive Bayes,decision trees,and random forests are also selected for comparison experiments.(3)Based on FastText's automatic scoring,the word vector is used to represent the text,and the word vector and the N-gram vector are superimposed to predict the category label of the word.The word vector uses a low-dimensional dense vector to represent the text,which can not only solve the dimensional explosion problem caused by the sparseness of the bag-of-words model,but also learn the position information of the words from the text,obtain the similarity between words,and solve the vocabulary gap problem.This paper verifies the feasibility of the text classification method on the graded protection evaluation data through experiments.Experiments show that using the TIFNG and TIF-NG-IG methods to extract features significantly improves the performance of the support vector machine,and the accuracy rates are increased by 4.3% and 4.8%,respectively.As far as the algorithm is concerned,the accuracy of FastText is 5.5% higher than that of the support vector machine,and the running time of FastText is much lower than that of the support vector machine.Comprehensive consideration,the FastTex model is more suitable for automatic scoring of iso-guarantee evaluation reports.