Anomaly-Based Intrusion Detection System Using Bidirectional Long Short-Term Memory for Internet of Things

The developing use of Internet of Things (IoT) applications in several aspects created a large number of data which is need the occurrence of existing techniques like fog and cloud computing. The Intrusion Detection System (IDS) is recognized several cybersecurity problems and its resources. In this paper, the Bidirectional Long Short-Term Memory (BiLSTM) is proposed for anomaly-based IDS in IoT network. The BoT-IoT and ToN-IoT datasets are used for identifying IDS in IoT networks. The normalization is used as preprocessing and Gain Ratio is used for selecting optimal features and the BiLSTM is used for classification of IDS in network which enables the model to scale large dataset and enhance the accuracy. The performance of BiLSTM is estimated by accuracy, f1 score, recall and precision. The BiLSTM achieves better accuracy of 98.76% and 96.84% for BoT-IoT and ToN-IoT dataset respectively which is better when compared to existing techniques like Group Theory Binary Spring Search-Hybrid Deep Neural Network (GTBSS-HDNN) and DNN.