A Three-layer Convolution Neural Network Approach for Intrusion Detection in IoT

The increasing number of connected devices in Internet of Things (IoT) has led to a rise in security threats and challenges in detecting intrusions in these environments. In this context, this research suggests an effective approach for IoT intrusion detection using a three-layer Convolution Neural Network (3L-CNN) approach. The proposed method comprises four major steps: data preprocessing, feature extraction, dimensionality reduction, and intrusion detection. The network traffic data collected from IoT devices is pre-processed using data cleaning, normalization, and balancing techniques. Feature extraction involves identifying statistical, frequency, time series, and protocol-specific features, which are then utilized to reduce the dimensionality of the data via an optimized autoencoder that uses self-adaptive Tasmanian Devil Optimization. Finally, a three-layer convolutional neural network approach is utilized to detect intrusions in the IoT environment. The proposed 3L-CNN approach achieves high accuracy, sensitivity, and specificity, as well as low false positive rates, in detecting network intrusions in IoT. The results demonstrate that the 3L-CNN approach is a powerful and effective technique for detecting intrusions in IoT environments and can adapt to different types of network traffic data.