计算机辅助定密系统关键技术研究与实现

发布时间：2018-10-09 15:57

【摘要】：计算机技术的飞速发展和信息化的不断推进,保密工作面临的问题日趋凸显,而定密工作是保密工作的源头性和基础性工作,决定了保密工作的效率和质量,因此,实现定密工作的高效性、准确性和智能性是现实的迫切需求。当前,国内各机关、单位主要采用人工定密的方式,存在着一些新老问题,比如,定密职责不规范、不明确;定密尺度难以把握、经验难以积累;涉密数据难以管理等问题。针对这些实际问题,本文针对文本定密分类关键技术开展研究,重点对CHI统计特征选择和TFIDF权重计算方法进行优化和改进,在此基础上设计并实现了融合规则定密和特征向量定密的计算机辅助定密系统,大大提升了定密工作效率。本文的工作主要如下:(1)针对数据集偏斜和文本篇幅差异,且传统特征选择方法和权重计算方法不能很好解决问题而导致分类不准确的情况,本文重点在CHI统计特征选择和TFIDF权重计算两个方面展开研究。定义了密度因子以修正原始数据,对特征分布进行优化,进而引入标准差和信息熵描述了特征类间分布和类内分布情况。在此基础上,将特征类内分布熵和词频因子融入CHI统计模型,改进了传统的CHI统计方法;综合了特征类间分布权和类内分布熵,改进了 TFIDF权重计算方法。针对两种方法,通过实验对准确率、召回率和F1值等性能指标进行了分析,实验表明,方法可以有效地提升文本分类的精度。(2)基于以上方法,并对涉密关键词库和定密规则库进行构建,设计并实现了计算机辅助定密系统。系统主要包括两部分功能,即基于规则库的文本定密和基于特征向量的文本定密方法。对系统实现进行了详细描述,并对系统的性能进行了测试,可较好提升定密工作的准确性和效率。
[Abstract]:With the rapid development of computer technology and the continuous promotion of information technology, the problems faced by security work are becoming increasingly prominent, and classified work is the source and basic work of security work, which determines the efficiency and quality of security work. It is an urgent need to realize the high efficiency, accuracy and intelligence of the secret work. At present, all organs and units in our country mainly adopt the method of manual secret, and there are some new and old problems, such as, the duty of keeping secret is not standardized and is not clear; the scale of determining secret is difficult to grasp, the experience is difficult to accumulate; and the data concerning secret is difficult to manage and so on. Aiming at these practical problems, this paper focuses on the research of key techniques of text classification, especially on the optimization and improvement of CHI statistical feature selection and TFIDF weight calculation method. On this basis, a computer-aided encryption system based on fusion rule and eigenvector encryption is designed and implemented, which greatly improves the efficiency of encryption. The main work of this paper is as follows: (1) aiming at the data set skew and the text length difference, and the traditional feature selection method and the weight calculation method can not solve the problem well, causes the classification inaccurate situation, This paper focuses on two aspects of CHI statistical feature selection and TFIDF weight calculation. The density factor is defined to modify the original data to optimize the feature distribution and then the standard deviation and information entropy are introduced to describe the distribution between and within the class. On this basis, the distribution entropy and word frequency factor in the feature class are incorporated into the CHI statistical model, the traditional CHI statistical method is improved, and the distribution weight among the feature classes and the intra-class distribution entropy are synthesized, and the calculation method of the TFIDF weight is improved. For the two methods, the accuracy, recall and F1 are analyzed through experiments. The experiments show that the method can effectively improve the accuracy of text classification. (2) based on the above methods, The secret keyword database and the secret rule base are constructed, and the computer aided encryption system is designed and implemented. The system mainly includes two functions: text encryption based on rule base and text encryption method based on feature vector. The implementation of the system is described in detail, and the performance of the system is tested, which can improve the accuracy and efficiency of the system.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1;D631.31

【相似文献】