A Novel Self-Supervised Learning-Based Anomalous Node Detection Method Based on an Autoencoder for Wireless Sensor Networks

Miao Ye

doi:10.1109/JSYST.2023.3347435

Return

A Novel Self-Supervised Learning-Based Anomalous Node Detection Method Based on an Autoencoder for Wireless Sensor Networks

2024-03-01

PRE

DOI：10.1109/JSYST.2023.3347435

Original

Original request for help

Save

Abstract

En 中文

To address the issue that existing wireless sensor network (WSN)-based anomaly detection methods only consider and analyze temporal features, in this article, a self-supervised learning-based anomalous node detection method based on an autoencoder is designed. This method integrates the extraction of temporal WSN data flow feature, spatial position feature, and intermodal WSN correlation feature into the autoencoder design to make full use of the spatial and temporal information of the WSN for anomaly detection. First, a fully connected network is used to extract the temporal features of nodes by considering a single mode from a local spatial perspective. Second, a graph neural network is used to introduce the WSN topology from a global spatial perspective for anomaly detection and to extract the spatial and temporal features of the data flows of nodes and their neighbors by considering a single mode. Then, an adaptive fusion method involving weighted summation is used to extract the relevant features between different modes. In addition, the gated recurrent unit network structure is introduced to solve the long-term dependence problem in the time dimension. Eventually, the reconstructed output of the decoder and the hidden layer representation of the autoencoder are fed into a fully connected network to calculate the anomaly probability for the current system. Since the spatial feature extraction operation is performed first, the designed method can be applied for anomaly detection in large-scale networks with the addition of a clustering operation. Experiments show that the designed method outperforms baseline methods, and the F1 score reaches 90.6%, at least 5.2% higher than those of existing anomaly detection methods based on unsupervised reconstruction and prediction.

Keywords:

Wireless sensor networks

Feature extraction

Anomaly detection

Data models

Correlation

Graph neural networks

Training

Fault detection

machine learning

neural networks

unsupervised learning