刘栋,蒋正威,朱英伟,黄银强,肖艳炜
LIU Dong,JIANG Zhengwei,ZHU Yingwei,HUANG Yinqiang,XIAO Yanwei

• 1:国网浙江省电力有限公司金华供电公司
• 2:国网浙江省电力有限公司

摘要(Abstract)：

随着智能电网的不断发展，电力系统网络安全问题日益凸显。为充分利用电网海量数据资源，提高大数据利用率，需要在流量数据中挖掘异常，加强电网抵抗网络安全风险的能力。利用深度学习的特点，对电网的电力监控系统网络流量进行快速检测，以保障电力系统数据采集和存储的安全。采用LSTM（长短期记忆）深度学习网络作为特征提取器，以解决数据特征提取困难的问题，提出了基于改进SVM（支持向量机）嵌入决策树模型的流量异常检测方法。从数据测试实验结果可知，该方法具有较高的准确率，优于多种传统方法。
With the continuous development of smart grid,network security of power system becomes increasingly prominent. In order to make full use of the massive data of power grid and improve the utilization rate of big data,it is required to mine the anomalies in massive traffic data to safeguard power grid by fending off network security risks. This paper,in the light of the characteristics of deep learning to quickly detect the network traffic of the power monitoring platform of the power grid,ensures the security of data acquisition and storage of the power system;besides,it uses LSTM(long short-term memory)deep learning network as the feature extractor to resolve the difficulty of data feature extraction;finally,the paper proposes a network traffic anomaly detection method based on decision tree model with SVM(support vector machine)embedded. It is concluded from the experimental results of data testing that the method is of high accuracy and superior to many traditional methods.

关键词(KeyWords)： 深度学习;特征提取;支持向量机;流量检测;网络安全
deep learning;feature extraction;support vector machine;traffic detection;network security

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司科技项目（5211JH1900M2）

作者(Author): 刘栋,蒋正威,朱英伟,黄银强,肖艳炜
LIU Dong,JIANG Zhengwei,ZHU Yingwei,HUANG Yinqiang,XIAO Yanwei

DOI: 10.19585/j.zjdl.202203011

参考文献(References)：

• [1]郭永亮.电力大数据背景下的电网规划研究[J].现代工业经济和信息化，2020,10(11):70-71.
• [2]刘曼琳，牛晓博.大数据面临的网络安全威胁及应对策略[J].网络安全技术与应用，2021(1):69-70.
• [3]赵菲.网络入侵检测中数据挖掘技术的应用研究[J].科技创新与生产力，2020(12):58-60.
• [4]杨至元，张仕鹏，孙浩.电力系统信息物理网络安全综合分析与风险研究[J].南方能源建设，2020,7(3):6-22.
• [5]王振东，张林，李大海.基于机器学习的物联网入侵检测系统综述[J].计算机工程与应用，2021,57(4):18-27.
• [6]曹磊，李占斌，杨永胜，等.基于双层注意力神经网络的入侵检测方法[J].计算机工程与应用，2021,57(19):142-149.
• [7]罗峰，胡强，侯硕，等.基于支持向量机的CAN-FD网络异常入侵检测[J].同济大学学报（自然科学版），2020,48(12):1790-1796.
• [8]ZARAI R.Recurrent neural networks&deep neural networks based on intrusion detection system[J].Open Access Library Journal,2020,7(3):1-11.
• [9]YANG A M,ZHUANSUN Y X,LIU C S,et al.Design of Intrusion Detection System for Internet of Things Based on Improved BP Neural Network[J].IEEE Access,2019,7:106043-106052.
• [10]HASSAN M M,Gumaei A,Alsanad A,et al.A hybrid deep learning model for efficient intrusion detection in big data environment[J].Information Sciences,2020,513:386-396.
• [11]许文华.基于深度神经网络的长文本分类算法的设计与实现[D].南京：南京邮电大学，2020.
• [12]隋金城.基于LSTM神经网络的股票预测研究[D].青岛：青岛科技大学，2020.
• [13]RANZATO F,ZANELLA M.Abstract interpretation of decision tree ensemble classifiers[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2020,34(4):5478-5486.
• [14]MAJI S,ARORA S.Decision tree algorithms for prediction of heart disease[M]//Information and communication technology for competitive strategies.Springer,Singapore,2019:447-454.
• [15]PAL M,FOODY G M.Feature selection for classification of hyperspectral data by SVM[J].IEEE Transactions on Geoscience and Remote Sensing,2010,48(5):2297-2307.
• [16]郭金玲.一种基于SVM的主动多分类方法[J].山西大学学报（自然科学版），2021,44(2):257-261.
• [17]张淑行，连晓峰，王焜.基于加权马氏距离多核学习的外观缺陷分类[J].传感器与微系统，2020,39(10):129-131.
• [18]YANG M,CHEN X,LUO Y,et al.An android malware detection model based on DT-SVM[J].Security and Communication Networks,2020,Article ID 8841233,11pages,2020.
• [19]张定华，胡祎波，曹国彦，等.面向工业网络通信安全的数据流特征分析[J].西北工业大学学报，2020,38(1):199-208.
• [20]石峻驿，周妮文，钱源.基于网络搜索数据的北京市旅游区客流量预测研究：基于五折交叉验证法的不同预测模型比较[J].经济统计学（季刊），2017(2):73-85.