Received:
Revised:
Online:
肖如良1,曾智霞2,肖晨凯1,张仕3
通讯作者:
Abstract: The development of Industrial Internet of Things (IIoT) systems has been widely used in the fields of security monitoring, intelligent transportation, environmental monitoring, etc. However, the sensors are damaged due to continuous use and normal wear and tear, resulting in hidden anomalies in the collected and recorded sensing data. The detection of anomalies in IIoT data is a challenging problem that is currently of great importance to both academia and industry. An anomaly detection algorithm (Locality Sensitive Bloom Filter (LSBF)) based on a locality sensitive Bloom Filter model is proposed. The proposed algorithm is based on a locality sensitive Bloom Filter model (LSBF), which uses the fast Johnson-Lindenstrauss transform based on spatial partitioning (SP-FJLT) to hash map the data, and then a Mutual Competition (MC) strategy to de-dry the data. The experimental results on dataset SIFT, MNIST and FMA show that the proposed algorithm exhibits high detection accuracy and low false alarm rate compared with the current mainstream anomaly detection algorithms, and can be effectively used for anomaly detection of IIoT data.
Key words: Industrial Internet of Things, anomaly detection, Bloom Filter, Fast Johnson-Lindenstrauss transform, Mutual Competition strategy, hidden anomaly
摘要: 工业物联网(IIoT)系统中传感器由于持续使用和正常磨损出现损坏,导致收集和记录的传感数据出现隐性异常。为解决该问题,提出一种基于局部敏感Bloom Filter模型的异常检测算法(LSBF)。首先利用基于空间划分的快速Johnson-Lindenstrauss变换(SP-FJLT)对数据进行哈希映射,然后采用相互竞争策略(MC)进行除噪,最后利用0-1编码构建Bloom Filter。在SIFT、MNIST和FMA三个基准数据集上进行仿真实验,实验结果表明,所提出的算法与当前主流的异常检测算法相比,表现出较高的检测准确率和较低的误报率,可有效应用于IIoT数据的异常检测。
关键词: 工业物联网, 异常检测, Bloom Filter, 快速Johnson-Lindenstrauss变换, 相互竞争策略, 隐性异常
CLC Number:
中图分类号:TP391
肖如良 曾智霞 肖晨凯 张仕. CCML2021+330: 基于局部敏感Bloom Filter的工业物联网隐性异常检测[J]. .
0 / Recommend
Add to citation manager EndNote|Ris|BibTeX
URL: http://www.joca.cn/EN/