Hybrid MQTTNet: An Intrusion Detection System Using Heuristic-Based Optimal Feature Integration and Hybrid Fuzzy with 1DCNN

Abstract

The development of IoT systems combined with the smart environments work in an effective way by making the objects to be smart. On the other hand, the IoT systems are vulnerable to different attacks based on security that may leads to considerable damage to IoT services. To meet this requirement, a novel intrusion detection system is proposed by using Message Queuing Telemetry Transport (MQTT) datasets. Initially, the IoT data is garnered from the standard MQTT datasets. The collected data is undergone for the pre-processing phase. Consequently, the pre-processed data is subjected to select the three set of features. The first features are chosen optimally by the Improved Vulture Starvation-based African Vultures Optimization Algorithm, the second features are acquired by the statistical features that are optimized by developed IVS-AVOA and the third features are obtained optimally by autoencoder with developed IVS-AVOA. Further, these resultant three features are fused with weight parameter to yield the weighted fused features. Finally, the classification is performed in the Hybrid Classifier, where it includes Fuzzy and One-Dimensional Convolutional Neural Network, in which the hyper parameters are tuned by improved AVOA algorithm. Hence, the proposed model demonstrates the effectiveness of defending the attack entailment in IoT environment.