随着工业互联网的进一步推进，船舶工业网络系统面临着更大的挑战。为进一步提高船舶工业网络的安全和稳定，网络入侵检测至关重要。本文提出基于深度学习的船舶工业网络入侵检测方法，采用字典数的方法对多种数据流量收集创建，利用针对船舶网络改进的深度学习算法进行船舶网络数据流的特征分层提取，并采用瀑布型融合方法将不同层的特征向量进行特征融合。利用softmax进行分类，划分为3个危险等级，在低危、中危、高危3种情况，实现网络入侵检测。现场测试结果表明，基于深度学习的船舶工业网络入侵检测模型的准确率较高，大大提高了网络入侵检测效率，弥补了传统技术无法检测未知入侵的弱点。
With the further development of industrial Internet, the network system of shipbuilding industry is facing greater challenges. In order to further improve the safety and stability of shipbuilding industry network, network intrusion detection is of great importance. In this paper, based on the deep study of the shipping industry network intrusion detection method, the number of dictionary method for a variety of data traffic collection is created, the depth of the learning algorithm was modified using in ship network layered extraction for shipping network data flow characteristics, and USES the waterfall fusion method to different layer of fusion, feature vector to use softmax classification, risk is divided into four levels: safety, low risk, moderate and high-risk four kinds of circumstances, the network intrusion detection. The field test results show that the deep learning based network intrusion detection model of shipbuilding industry has a higher accuracy, greatly improves the efficiency of network intrusion detection, and makes up for the weakness of traditional technology that cannot detect unknown intrusion.
2020,42(5): 181-183 收稿日期：2019-03-06
DOI：10.3404/j.issn.1672-7649.2020.05.034
分类号：U661.43
作者简介：朱军(1987-),男,高级工程师,从事工业网络安全、人工智能研究工作
参考文献：
[1] 李聪, 贾红军. 船舶网络入侵风险等级估算研究[J]. 舰船科学技术, 2018, v. 40(12): 56–58
[2] 章昱, 李腊元. 网络入侵检测系统的设计与实现[J]. 武汉理工大学学报(交通科学与工程版), 2004, 28(5): 657–660
[3] 李辉, 管晓宏, 昝鑫, 等. 基于支持向量机的网络入侵检测[J]. 计算机研究与发展, 2003, 40(6): 799–807
[4] 支持向量机算法及其在网络入侵检测中的应用[D]. 大连: 大连海事大学, 2004.
[5] 尹宝才, 王文通, 王立春. 深度学习研究综述[J]. 北京工业大学学报, 2015(1): 48–59
