Computer Network Security Optimization Model Based on Improved Neural Network Algorithm

The speed and efficiency of cyber-attacks have recently increased the great momentum and much importance of computer network security. An effective safety model will be essential to protect sensitive data, and for network integrity, several algorithms are proposed; thus, the research provides an efficient optimization method on computer network security by constructing an improved neural network algorithm. The research introduces the Enriched Marine Predators algorithm-tuned Improved Convolutional Neural Network (EMP-ICNN), a hybrid model that improves the accuracy in detecting network intrusions and aims to construct a general security optimization framework to classify and prevent malicious attacks in an autonomous manner. The results of the proposed model evaluation are presented by using open-source network traffic datasets, representative of both data training and testing purposes. Normalization techniques have been used for pre-processing the data to scale the input appropriately to the model. Kernel Principal Component Analysis (Kernel-PCA)-based feature extraction has been conducted to reduce dimensionality, emphasizing those features most relevant to contribute to the attack detection process. This EMP-ICNN contains an enriched optimization approach in which the optimization of architecture via the EMP algorithm was conducted in such a way that this reduces false positives and, in turn, increases accuracy in the detection of attacks. The model proposed is implemented in Python software, showing very good accuracy of 96.7%, precision of 92.6%, recall of 94.6%, and F1 score of 93.3% in the classification of attacks. Especially, it outperforms conventional models with much higher values in all metrics. Therefore, the proposed model has shown great enhancement on computer network security, hence carrying a very promising solution toward advanced security applications.