A Density-based Clustering and Deep Learning Algorithm for Intrusion Detection in Sensor Networks

Abstract

At present, network security is leading a dynamic role in wireless sensor networks and security becomes a precarious challenge in wireless sensor networks. Therefore, an Intrusion detection system (IDS) is a secure mechanism, which is aimed to identify and prevent from unapproved access, as it maintains harmless and protects network systems. Keeping these limitations, we present an approach called DBDNN, which combines Density-based spatial clustering of applications with noise (DBSCANSC) and the deep neural network (DNN). First, “split the given dataset into subsets depend on similarity features by core point”, as in DBSCANSC. Secondly, “the distance between data points in training dataset and testing dataset calculated by using closely reachable points and which is fed input to deep neural network system”. This study used KDD-Cupp99 datasets to check the implementation of the model. The experimental results indicate that the proposed DBSCANSC-DNN performs higher than Bayesian classifier (Bayes), Backpropagation Neural Networks (BPNN), Spectral Clustering and Deep Neural Networks (SCDNN) and the Support Vector Machine (SVM). Finally, the proposed method provides an effective technique for analysis of IDS in huge networks of anomalous attacks detection.