A Hybrid Classification Approach for Intrusion Detection in IoT Network

Choudhary, Sarika ; Kesswani, Nishtha


With the increase in number of IoT devices, the capabilities to provide reliable security and detect the malicious activities within the IoT network have become quite challenging. We propose a hybrid classification approach to detect multi-class attacks in the IoT network. In the proposed model, Principle Component Analysis (PCA) is used to extract the useful features and Linear Discriminant Analysis (LDA) is used to reduce the high dimension data set into lower dimension space by keeping less number of important features. This was assisted by use of a combination of neural network and Support Vector Machine (SVM) classifiers to improve the detection rate and decrease the false alarm rate. The neural network, a multi-class classifier, is used to classify the intruders in the network with more accuracy. The SVM is an efficient and fast learner classifier which is used to classify the unmatched behavior. The proposed method needs less computation complexity for intrusion detection. The performance of the proposed model was evaluated on two benchmark datasets for intrusion detection, i.e., NSL-KDD and UNSW-NB15. Results show that our model outperforms existing models.


Internet of Things (IoT), Intrusion Detection Systems (IDS), Linear Discriminant Analysis (LDA), PCA, SVM

Full Text: PDF (downloaded 361 times)


  • There are currently no refbacks.
This abstract viewed 448 times