Title: Real time detection of intrusion trace information in sensor network based on Bayesian belief network

Authors: Hongli Deng; Tao Yang

Addresses: Education and Information Technology Center, China West Normal University, Nanchong 637002, China ' Education and Information Technology Center, China West Normal University, Nanchong 637002, China

Abstract: In order to overcome the singularity of intrusion detection in sensor networks, a real-time detection method of intrusion trace information in sensor networks based on Bayesian belief network is proposed. This method constructs the intrusion detection model of Bayesian belief network based on the directed bipartite graph, uses the improved wavelet threshold to denoise the sensor network signal, and extracts the trace information features with abnormal conditions in the denoised signal. The most representative abnormal trace information feature is extracted by principal component analysis (PCA). Based on this feature, Bayesian belief network intrusion detection model is used to realise the real-time detection of trace information of sensor network intrusion. The experimental results show that the overall detection rate is higher than 90%, the detection accuracy is higher; the recall rate and precision rate are both higher than the traditional method, and the maximum error of the detection results is only 0.05.

Keywords: Bayesian belief; network; sensor network; intrusion trace; real-time detection; directed bipartite graph; denoised signal.

DOI: 10.1504/IJAACS.2023.129644

International Journal of Autonomous and Adaptive Communications Systems, 2023 Vol.16 No.1, pp.48 - 65

Received: 17 Apr 2020
Accepted: 01 Sep 2020
Published online: 17 Mar 2023 *