A Novel Graph Structure Learning Based Semi-Supervised Framework for Anomaly Identification in Fluctuating IoT Environment

作     者：Weijian Song Xi Li Peng Chen Juan Chen Jianhua Ren Yunni Xia 

作者机构：School of Computer and Software EngineeringXihua UniversityChengdu610039China West China Second University HospitalSichuan UniversityChengdu610065China School of Computer ScienceChongqing UniversityChongqing400044China 

出 版 物：《Computer Modeling in Engineering & Sciences》 (工程与科学中的计算机建模（英文）)

年 卷 期：2024年第140卷第9期

页      面：3001-3016页

核心收录：

学科分类：0809[工学-电子科学与技术（可授工学、理学学位）] 08[工学] 0701[理学-数学] 

基　　金：This research is partially supported by the National Natural Science Foundation of China under Grant No.62376043 Science and Technology Program of Sichuan Province under Grant Nos.2020JDRC0067,2023JDRC0087,and 24NSFTD0025 

主　　题：IoT multivariate time series anomaly detection graph learning semi-supervised mean teachers 

摘      要：With the rapid development of Internet of Things(IoT)technology,IoT systems have been widely applied in health-care,transportation,home,and other ***,with the continuous expansion of the scale and increasing complexity of IoT systems,the stability and security issues of IoT systems have become increasingly ***,it is crucial to detect anomalies in the collected IoT time series from various ***,deep learning models have been leveraged for IoT anomaly ***,owing to the challenges associated with data labeling,most IoT anomaly detection methods resort to unsupervised learning ***,the absence of accurate abnormal information in unsupervised learning methods limits their *** address these problems,we propose AS-GCN-MTM,an adaptive structural Graph Convolutional Networks(GCN)-based framework using a mean-teacher mechanism(AS-GCN-MTM)for anomaly *** performs better than unsupervised methods using only a small amount of labeled *** Teachers is an effective semi-supervised learning method that utilizes unlabeled data for training to improve the generalization ability and performance of the ***,the dependencies between data are often unknown in time series *** solve this problem,we designed a graph structure adaptive learning layer based on neural networks,which can automatically learn the graph structure from time series *** not only better captures the relationships between nodes but also enhances the model’s performance by augmenting key *** have demonstrated that our method improves the baseline model with the highest F1 value by 10.4%,36.1%,and 5.6%,respectively,on three real datasets with a 10%data labeling rate.