Hybrid deep learning-based intrusion detection system for wireless sensor network

Abstract

Wireless Sensor Networks (WSNs) play an important role in the modern era and security has become an important research area. Intrusion Detection System (IDS) improve network security by monitoring the network state so that threats and attacks can be detected and rectified. With decades of development, IDS still lags in performance in terms of detection accuracy, false alarm rate and unknown attack detection. To overcome this performance issue, researchers implemented numerous machine learning techniques to detect the attacks. Conventional machine learning models identify the essential features through specific feature extraction techniques which increases the computation complexity of the system. For attack details and their sub-categories, deep learning technique is used in the proposed work. The detection model incorporates ResNet based on Inception with a support vector machine to detect WSN intrusions. Proposed algorithm is applied to Standard NSL-KDD data set and performance metrics like recall, precision, accuracy and f1-score are considered for analysis. The comparative analysis demonstrates the proposed model performance of 99.46% accuracy is better than traditional approaches like random forest, decision tree, deep neural network and convolutional neural network.

Topics & Concepts