Journal
IEEE ACCESS
Volume 7, Issue -, Pages 96319-96333Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/ACCESS.2019.2929581
Keywords
Outlier detection; wireless sensor networks (WSN); iforest; local outlier factor (LOF); isolation using nearest neighbor ensembles (iNNE); sliding window
Categories
Funding
- Shaanxi Science and Technology Coordination and Innovation Project [2016KTZDGY04-01]
- International Science and Technology Cooperation Program of the Science and Technology Department of Shaanxi Province, China [2018KW-049]
- Special Scientific Research Program of Education Department of Shaanxi Province, China [17JK0711]
- Communication Soft Science Program of Ministry of Industry and Information Technology, China [2019-R-29]
Ask authors/readers for more resources
In recent years, wireless sensor networks have been extensively deployed to collect various data. Due to the effect of harsh environments and the limitation of the computing and communication capabilities of sensor nodes, the quality and reliability of sensor data are affected by outliers. Thus, an effective outlier detection method is essential. The existing outlier detection methods have some drawbacks, such as extra resource consumption introduced by the size growth of a local detector, poor performance of combination methods of local detectors, and the weak adaptability of the dynamic changes of the environment, etc. We propose an isolation-based distributed outlier detection framework using nearest-neighbor ensembles (iNNE) to effectively detect outliers in wireless sensor networks. In our proposed framework, local detectors are constructed in each node by the iNNE algorithm. A new combination method taking advantage of the spatial correlation among sensor nodes for local detectors is presented. The method is based on the weighted voting idea. In addition, we introduce a sliding window to update local detectors, which enables the adaption of dynamic changes in the environment. The extensive experiments are conducted on two classic real sensor datasets. The experimental results show our framework significantly improves the detection accuracy and reduces the false alarm rate compared with other outlier detection frameworks.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available