IoT intrusion detection system using ensemble classifier and hyperparameter optimization using tuna search algorithm

Abstract

The Internet of Things (IoT) is a dynamic and delightful research field in this emerging technology. It can be globally connected with many IoT devices and exchange a large amount of data. However, the threats also developed and misguided the entire network’s behaviour. This article proposes an Intrusion Detection System (IDS) using the proposed ensemble classifier along with the Tuna Swarm Optimization (TSO) to fine-tune the hyperparameters and help to enhance the detection accuracy of attacks that take place in IoT environment. Here, the publicly available message queue telemetry transport (MQTT) network dataset is used to classify the given data into the following categories: SlowlTe, malformed, brute force, flood, DoS, and legitimate. Initially, the dataset is pre-processed to remove possible outliers, then data balancing is performed using the Synthetic Minority Oversampling Technique (SMOTE) technique and features are extracted with the help of Recursive Feature Elimination (RFE). Finally, ensemble classifier along with the optimized parameters using TSO helps in detecting the attacks in IoT attacks. The proposed TSO-ensemble classifier achieved a classification accuracy of 99.12%. In contrast, the classification accuracy of the existing Improved Vulture Starvation-based African Vultures Optimization (IVS-AVOA) and Convolutional Neural Network Long Short-Term Memory (CNN-LSTM) have achieved a classification accuracy of 96.61% and 98.94% respectively.