An intrusion detection model using improved convolutional deep belief networks for wireless sensor networks

Intrusion detection is a critical issue in the wireless sensor networks (WSNs), specifically for security applications. In literature, many classification algorithms have been applied to address the intrusion detection problems. However, their efficiency and scalability still need to be improved. This paper proposes an improved convolutional deep belief network-based intrusion detection model (ICDBN_IDM), which consists of a redundancy detection algorithm based on the convolutional deep belief network and a performance evaluation strategy. The redundancy detection can remove non-effective nodes and data, and save the energy consumption of the whole network. The improved algorithm extracts features from normal and abnormal behaviour samples by using unsupervised learning and overcomes the problem of unknown or less prior samples. Compared with the commonly used machine learning mechanisms, the proposed ICDBN_IDM achieves high intrusion detection accuracy, reduces the ratio of the false alarm while saving the energy consumption of sensor nodes.

Automated summary

The study introduces an intrusion detection model based on convolutional deep belief networks, which effectively addresses issues of redundancy and energy consumption in intrusion detection. By utilizing unsupervised learning for feature extraction, the model enhances intrusion detection accuracy and reduces false alarm rates.