A Smart Intrusion Detection Networked System Against Cyber Attacks Based on Intelligent Algorithms

Abstract

The Internet of Things (IoT) faces growing security challenges, with escalating threats targeting devices, necessitating the development of efficient, secure intrusion detection systems (IDS). Deep learning (DL) and machine learning (ML) offer promising solutions to enhance detection accuracy and network security in IoT environments. In this paper, we propose a hybrid deep learning classification model (DNN-LSTM), applied to the Multi-Step Cyber–Attack Dataset (MSCAD), which consists of 66 features. We apply the DNN (deep neural network), LSTM (long short-term memory), and hybrid model (DNN-LSTM) to the dataset to evaluate their performance.
The experimental findings showcase that our proposed model surpasses the accuracies achieved by DNN, LSTM models, and hybrid models, with scores of 99.72%, 99.76%, and 99.78%, respectively. It demonstrates their exceptional ability to produce the most accurate predictions. The results highlight the superiority of the DNN-LSTM hybrid model in identifying multi-step cyberattacks, reducing false positives, and enhancing the overall efficiency of intrusion detection in IoT systems. Our work presents a promising approach to strengthening security in IoT networks and contributes to the advancement of intrusion detection systems through the use of advanced deep learning techniques. These findings indicate that hybrid models like DNN-LSTM can play a critical role in addressing the evolving security challenges of IoT.