深度学习算法的船舶通信网络非法入侵行为识别研究  

Research on the identification of illegal intrusion in ship communication network based on deep learning algorithm

作　　者：韩睿[1] HAN Rui(Shanxi Polytechnic College School,Taiyuan 030006,China)

机构地区：[1]山西职业技术学院,山西太原030006

出　　处：《舰船科学技术》2020年第6期163-165,共3页Ship Science and Technology

摘　　要：非法入侵严重影响船舶通信网络安全运行,船舶通信网络非法入侵行为具有很强的变异行为,导致当前船舶通信网络非法入侵行为的识别效果差。为了对各种船舶通信网络非法入侵行为进行准确性识别,提出深度学习算法的船舶通信网络非法入侵行为识别技术。该技术将船舶通信网络非法入侵行为识别看作是一个模式分类问题,将非法入侵行为划分多种类型,然后提取各种船舶通信网络非法入侵行为的变化特征,采用深度学习算法对变化特征和船舶通信网络非法入侵行为类型之间的联系进行分析,以区别各种船舶通信网络非法入侵行为,最后选择有代表性的船舶通信网络非法入侵行为进行了性能测试。结果表明,深度学习算法的船舶通信网络非法入侵行为识别率高于95%,非法入侵行为识别时间控制在2 s以内,可以满足现代船舶通信网络通信安全的需要。The illegal intrusion seriously affects the safe operation of the ship communication network, and the illegal intrusion behavior of the ship communication network has a strong mutation behavior, which leads to the poor recognition effect of the current illegal intrusion behavior of the ship communication network. In order to accurately identify the illegal intrusion behavior of various ship communication networks, a deep learning algorithm is proposed to identify the illegal intrusion behavior of the ship communication network Technology. This technology regards the identification of the illegal intrusion of ship communication network as a pattern classification problem, divides the illegal intrusion into many types, then extracts the changing characteristics of various illegal intrusion of ship communication network, and analyzes the relationship between the changing characteristics and the types of illegal intrusion of ship communication network by using the deep learning algorithm, so as to distinguish various kinds of ship communication Finally, we choose the representative ship communication network illegal intrusion behavior for performance test. The recognition rate of the deep learning algorithm is higher than 95%. The recognition time of an illegal intrusion is controlled within 2 seconds, which can meet the needs of modern ship communication network communication security.

关 键 词：深度学习 非法入侵 行为识别 多分类问题 现代通信技术 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]