MDNET: A Novel Neural Network Based on CNN and Fuzzy Rough Set with Adaptive Parameters for Intrusion Detection in the Internet of Things

The internet of things (IoT) has become increasingly widespread, finding applications in various fields like transportation fleet management, smart homes, traffic management, and smart greenhouses. However, a significant concern with IoT devices is the vast amount of data they store. Sensitive data, including personal information, requires strong security, as unauthorized access can lead to cyberattacks. Attack scenarios, including data theft, system hijacking, and unauthorized access, pose significant threats, making early detection crucial to minimize potential damage. This paper proposes the MDNET neural network for enhanced IoT intrusion detection. The MDNET network aims to achieve high accuracy in detecting IoT network intrusions without requiring manual hyperparameter tuning. To achieve this, the CEO Optimization Algorithm was employed to adaptively select the hyperparameters for the proposed network. Our proposed method employs fuzzy rough set theory to select appropriate features from the network traffic dataset. This process removes features significantly impacted by noise, thereby enhancing the overall efficiency of the network.The proposed approach was evaluated using five standard metrics accuracy, average precision, Kappa coefficient, Hamming loss, and Jaccard similarity; compared with seven existing models (FWP-SVM-GA, FBiNN, GA-FR-CNN, APSO-CNN, LSTM, GRU, ANN). The proposed method achieved an impressive ۹۹.۵۷% accuracy, ۹۹.۷% average-precision, and ۰.۹۹ Kappa coefficient, significantly outperforming the best-performing existing algorithm (FBiNN) with ۹۷.۲۸% accuracy and ۰.۹۶ Kappa coefficient.  Additionally, MDNET exhibited lower Hamming loss (۰.۰۰۳) and a higher Jaccard similarity (۰.۹۹), confirming its superior classification accuracy and robustness in IoT attack detection.

کلیدواژه ها:

Novel Neural Network ، Intrusion Detection ، Fuzzy Rough Set ، Chaotic Equilibrium Optimizer Algorithm ، Internet of Things network security ، Classify type attack