An Ensemble Intrusion Detection System based on Acute Feature Selection

As the Internet of Things (IoT), 5G, and Artificial intelligence (AI) continue to converge, the number of security incidents and occurrences on the networks has recently increased. Since more devices are connected to IoT networks, security is becoming a major concern. Conventional intrusion detection systems (IDS) are not well suited for use in the complex lightweight IoT environment. This research paper presented an IDS for the smart city environment based on IoT- Message queuing telemetry transport (MQTT) networks that could detect attacks using shallow learning algorithms. The proposed framework has four parts (i) a smart city network model with multiple MQTT clients (sensors and IoT devices) is created with the help of hardware. (ii) Injected a flooding attack on the MQTT broker to create the IDS dataset with normal and attack features, (iii) Based on the acute feature selection algorithm to select the optmized features from the raw dataset and validated with the Jaccard coefficient. (iv) The dataset is further trained and validated using shallow learning algorithms such as extreme gradient boosting (XGB), K-nearest Neighbors (KNN) and Random forest (RF). Experimental results outperform with better attack detection rate, attack prediction rate and improved accuracy over 97% with lower redundancy using selected features. Experimental results show that the proposed approach is more vulnerable to attacks in the IoT network.