关键词: 入侵检测, 无线传感器网络, 随机森林, 深度森林算法, 集成分类器

Abstract: In view of the poor performance, poor real-time detection along with higher model complexity of the existing feature selection algorithm and classification algorithm adopted in the wireless sensor network (WSN) intrusion detection system, a distributed WSN intrusion detection model based on the random forest algorithm and the deep forest algorithm is presented in this paper. The proposed model preprocesses traffic data passing through all sensor nodes first, and then allocates lightweight random forest classifier to sensor nodes and cluster head nodes respectively. Sensor nodes and cluster head nodes work up together to deal with the traffic data. Consequently, with the help of the deep forest algorithm emerged attacking behaviors can be seized at a base station from a large amount of traffic data. Intrusion detection can be thus carried out by means real-time classification of intrusion behaviors in WSN. In order to validate effectiveness of the proposed model, we use WSN-DS and NSL-KDD data sets to evaluate the reliability of the proposed model. The experimental results show that, comparing with the existing intrusion detection model, the proposed model does having better detection performance and better real-time performance. The weakness of over fitting in modeling can be avoided as a result.

Key words: intrusion detection, wireless sensor network (WSN), random forest, deep forest algorithm, ensemble classifier