基于文本的网络安全事件检测技术研究与探索  

Research and exploration of text-based network security incident detection technology

作　　者：周岳亮[1] 李宗阳[1] 李绍铭 ZHOU Yueliang;LI Zongyang;LI Shaoming(Affiliated Hospital of Youjiang Medical University for Nationalities,Baise Guangxi 533000,China)

机构地区：[1]右江民族医学院附属医院,广西百色533000

出　　处：《自动化与仪器仪表》2023年第8期79-82,共4页Automation & Instrumentation

基　　金：百色市科学研究与技术开发计划项目(百科20182504)。

摘　　要：针对基于文本的网络安全事件检测技术进行了研究与探索,以医院机密邮件泄露事件为例,设计了一个基于Spark的医院机密邮件快速检测系统,并对其关键模块即文本内容检测部分进行了研究与设计。首先,对医院机密邮件快速检测系统的整体框架进行了设计;然后提出了联合基于BLEU的文本词语相似度检测与基于BERT-BiLSTM单句语义相似度检测的方法对医院邮件文本内容进行检测;最后,对经过医院网络传输的邮件进行采集并建立数据集,分别对单个检测方法与联合检测方法进行实验,并对医院机密邮件快速检测系统进行测试,实验测试结果表明:联合检测方法大幅提高了检测的精确度与稳定性,设计的医院机密邮件快速检测系统能够对经由医院网络发送的邮件进行安全检测,其对普通邮件的检测准确率为97.78%,对含有机密信息邮件的检测准确率为94.66%。Taking the hospital confidential email leakage incident as an example,this paper designs a rapid detection system for hospital confidential email based on Spark,and studies and designs its key module,namely the text content detection part.Firstly,the overall framework of the hospital confidential mail rapid detection system is designed.Then,a method of combining BLEU-based text word similarity detection and BERT-BiLSTM-based single sentence semantic similarity detection is proposed to detect the text content of hospital emails.Finally,the mail transmitted through the hospital network is collected and a data set is established,the single detection method and the joint detection method are experimented,and the hospital confidential mail rapid detection system is tested,the experimental test results show that the joint detection method greatly improves the accuracy and stability of detection,and the hospital confidential mail rapid detection system designed in this paper can perform security detection on the mail sent through the hospital network,and its detection accuracy of ordinary mail is 97.78%.The detection accuracy rate of emails containing confidential information was 94.66%.

关 键 词：网络安全 文本检测 BLEU BERT-BiLSTM 

分 类 号：TP14[自动化与计算机技术—控制理论与控制工程]