Threat analysis model to control IoT network routing attacks through deep learning approach

Abstract

Most of the recent research has focused on the Internet of Things (IoT) and its applications. The open interface and network connectivity of the interconnected systems under the IoT network make them vulnerable to hackers. A model has been proposed to identify and classify IoT routing attacks. To generate IoT routing datasets, the Cooja simulator is used at first. The IoT routing dataset is then augmented into larger volumes using ADASYN, which is also used to solve the class imbalance problems. A deep learning hybrid model based on a Long-Short-Term Memory (LSTM) network and adaptive Mayfly Optimization Algorithm (LAMOA) was presented for the classification of IoT attacks. The adaptive MOA adjusts the weights in the various layers of the LSTM network and the Fully Connected Layer with SoftMax Classification. As part of the validation process, the proposed model was also compared with benchmark datasets NSL-KDD, BoT-IoT, and IoT-23. Using benchmark datasets, LAMOA achieved 99.94% accuracy for multiclassifications, 99.92% accuracy for binary classifications, and 98.42% accuracy for real-time datasets. Compared to other models, our proposed model significantly improves the accuracy of each attack's classification by 5–10%.