摘要
网络行为识别一直是网络安全中的研究热点,随着网络中数据量的海量增大以及数据的非线性等问题的影响,对于网络行为识别的特征提取和识别技术提出更高的要求。文章提出了一种基于堆叠自动编码器的网络行为识别方法,该方法通过构建堆叠自动编码器和SOFTMAX分类器的深度学习框架,结合无监督的预训练和有监督的全局微调,优化堆叠自动编码器的特征提取性能,实现了网络行为特征的深度提取,从而对高校流量数据中上网行为进行分析识别。
Network behavior recognition is always a research hotspot in network security.With the increase of the mass of data in the network and the influence of the nonlinear data,a higher requirement for the feature extraction and recognition technology of network behavior recognition is presented.Therefore,a network behavior recognition method based on stacked automatic encoder is proposed.The method constructs a deep learning framework of a stacked automatic encoder and a SOFTMAX classifier and combines unsupervised pre-training and supervised global fine-tuning to optimize feature extraction performance of the stacked automatic encoder and achieve deep extraction of network behavior features.In this way,the online behavior in the university traffic data is analyzed and identified.
作者
刘任熊
田由辉
张朝龙
LIU Renxiong;TIAN Youhui;ZHANG Chaolong(Network Information Center,Jiangsu Vocational Institute of Commerce,Nanjing 211168,China;Jiangsu Vocational Data Center,Nanjing 211168,China;School of Electric Engineering and Automation,Hefei University of Technology,Hefei 230009,China)
出处
《合肥工业大学学报（自然科学版）》
CAS
北大核心
2019年第2期189-194,共6页
Journal of Hefei University of Technology：Natural Science
基金
国家重点研发计划资助项目(2016YFF0102200)
国家自然科学基金资助项目(51607004
51577046)
国家自然科学基金重点资助项目(51637004)
江苏省教育科学规划资助项目(D/2016/03/54)
江苏省高等教育教改资助项目(2017JSJG308)
关键词
网络行为
识别
特征提取
深度学习
堆叠自动编码器
network behavior
recognition
feature extraction
deep learning
stacked automatic encoder
