Scaled IoT Intrusion Detection Model based on Improved PSO Algorithm Optimization

An improved convolutional nveural network intrusion detection training model is proposed to address the issues of slow convergence and low accuracy that may arise from the large randomness of initial parameters in traditional neural network intrusion detection models. The proposed approach utilizes an enhanced particle swarm optimization (PSO) algorithm to optimize the convolutional neural network intrusion detection model, thereby improving its global search capability. The improved PSO algorithm is employed to optimize the neural network hyperparameters, with the loss function used as the population location. Once the optimal parameters are obtained, a scaled convolutional neural network is constructed, and the model is trained using backpropagation. The initial model defaults to using the Gram's corner field method to transform the data into three-dimensional format, incorporating a softpool pooling layer to enhance feature extraction. Additionally, a scaled neural network structure is employed to maximize the perceptual field, and the CUDA parallel computing method is utilized to accelerate computation speed. Experimental results demonstrate that the proposed method exhibits certain advantages over the traditional PSO algorithm in terms of optimizing the neural network intrusion detection training method, as evaluated comprehensively in terms of efficiency and accuracy. These findings highlight the potential of the proposed method for improving the efficiency and accuracy of neural network intrusion detection training when compared to the traditional PSO algorithm.