摘要
在基于物联网的粮情监控系统中,传统入侵检测方法很难准确实时地从海量数据提取特征信息来识别网络攻击,使系统容易遭受安全问题,从而破坏数据的完整性,因此,提出一种基于深度信念网络的交替决策树入侵检测(DBNADT)方法。该方法利用深度信念网络进行无监督地特征学习,通过预训练将原始数据特征降维,利用权值微调算法获得数据的最优低维表示;然后采用交替决策树分类器对各种异常网络入侵数据进行识别。结果表明,DBN-ADT方法在攻击行为的识别准确率上比支持向量机(SVM)和逻辑回归(LR)分别提高了7.24%和8.25%,在检测时间上分别缩短了约1/2和2/5。DBN-ADT方法具有更高的检测准确率和实时性。
It is difficult for the food monitoring system based on the Internet of Things to extract feature information from massive data in real time to identify network attacks,making the system vulnerable to security problems and thus destroying data integrity.So an alternating decision tree intrusion detection method based on deep belief network(DBN-ADT)is proposed.The method uses the deep belief network to perform unsupervised feature learning.The pre-training transforms the original data into features,and uses the weight fine-tuning algorithm to obtain the optimal low-dimensional representation of the data.Then,the alternating decision tree classifier is used to identify various anomaly network intrusion data.The experiment results show that the DBN-ADT method improves the recognition accuracy of the attack behavior by 7.24%and 8.25%,and shortens the detection time by about 1/2 and 2/5,respectively.compared with the support vector machine(SVM)and the logistic regression(LR),The DBN-ADT method has higher detection accuracy and better real-time performance.
作者
宋雪桦
汪盼
邓壮来
解晖
SONG Xuehua;WANG Pan;DENG Zhuanglai;XIE Hui(School of Computer Science and Communication Engineering,Jiangsu University,Zhenjiang 212000,China)
出处
《安庆师范大学学报（自然科学版）》
2020年第1期1-8,14,共9页
Journal of Anqing Normal University(Natural Science Edition)
基金
国家重点研发计划（2017YFC1600804）
关键词
物联网
入侵检测
深度信念网络
交替决策树
粮情监控系统
Internet of Things
intrusion detection
deep belief network
alternating decision trees
food monitoring system
