Dingo optimized deep learning framework for intrusion detection in wireless sensor network

Abstract

Wireless sensor networks (WSNs) are becoming more and more widespread in today's world. WSNs are widely employed in various applications, including military, industrial, healthcare, smart homes, and smart cities. All WSN applications demand reliable communication between sensor nodes and base stations. Some of these applications deal with private and critical data that needs to be secured from unauthorized access. Unfortunately, weaknesses in WSN design and the limited processing power of individual nodes make them vulnerable to a range of attacks. This work proposes a novel Dingo Optimized Deep Learning for Intrusion Detection (DOLI) technique to detect intrusion in WSNs. Initially, the data from the sensor nodes are pre-processed and classified using the BiLSTM model which classifies the outputs into attack and non-attack. The effectiveness of the suggested approach has been assessed with False Alarm Rate (FAR), Detection Rate (DR), F1-Score, recall, accuracy, precision, and time consumption. The proposed DOLI method attains an accuracy of 98.73% in the KDD Cup99 dataset whereas the existing FFDNN, CDBN, and CRF-LFS methods achieve an accuracy of 76.64%, 75.59%, and 80.39% respectively.