Intelligent cyberattacks causes significant risks to data security, requiring advanced network protection mechanisms. While traditional methods like firewalls, authentication, and antivirus programs offer some defence, they often fail to ensure complete security. Intrusion detection systems (IDS) provide an additional layer of protection by dynamically identifying cyberattacks. This study focuses on network intrusion detection using a hybrid filter-based approach with neural networks. The proposed method consists of three key modules: segmentation and feature selection, training a neural network using filtered features, and testing the trained model. Feature selection is performed using the shuffled frog leaping algorithm (SFLA), enabling the identification of the most relevant optimal subset of features for classification. To achieve enhanced performance, the training employs an error back-propagation neural network. By combining SFLA-based feature selection with deep learning techniques, this work provides an effective framework for enhancing IDS performance through an optimal feature subset on NSL-KDD and KDD Cup'99 datasets.