Abstract
The intrusion detection in the Internet of Things (IoT) network presents a number of difficulties, necessitating the skillful use of network device attributes for precise threat identification. With a multi-phase approach, this research provides a novel intrusion detection method. This technique takes advantage of the non-linear relationship identification capability of the Chatterjee correlation. The Artificial Bee Colony (ABC) approach is employed for optimizing the feature selection procedure. Lastly, a specialized Residual Neural Network (ResNet) is designed that can detect complex relationships from the data using minimal computational cost. The above design contains three types of residual blocks which are designed in a discriminatory fashion to enable the model to capture both the high-level and low-level features. At each layer, an increase in the number of filters is used to enhance the process of feature extraction. The convolutional layers are of great importance since they control the whole learning process. To ensure the effectiveness of the network in collecting complex relations, a method for converting 1D features into 2D images is provided to facilitate the ResNet input. Experimental results indicate that the suggested approach is exceptionally adept at identifying intrusions in the IoT networks, achieving accuracy of 98.5711%. The combination of ABC optimization, ResNet, and Chatterjee correlation leads in a strong and effective intrusion detection system, which shows promise for improving the cyber security of IoT environments.