Over the past few years, the increasing use of computer networks and binary information transmission has highlighted the importance of information security and intrusion detection in computer systems and networks. Simultaneously with the introduction of new and diverse attacks, various methods and systems in intrusion detection systems are proposed. An intrusion detection system is essentially a set of tools, methods, and documentation to detect and report unauthorized network activity. Static techniques, clustering, and learning can be introduced as the most common intrusion detection methods. One of the well-known models in intrusion detection is modeling based on neural networks. This model has shown outstanding performance in various fields, from forecasting to diagnosis and classification. One of the challenges of neural networks is the regulating parameters by an expert. This study aims to improve neural network performance by providing a hybrid model for automatic regulating of network parameters. The NSL-KDD dataset is considered a modified version of KDD-CUP99 to test and evaluate the proposed model. The results proved that the proposed model has a better learning capability than the conventional neural networks.
