Processed dataset to train and test the WGAN-GP_IMOA_DA_Ensemble model

In the dynamic landscape of cybersecurity, robust and efficient Intrusion Detection Systems (IDS) are essential. However, traditional IDS models often struggle with high-dimensional features, imbalanced datasets, and evolving attack patterns. To overcome these limitations, we propose a hybrid framework named WGAN-GP_IMOA_DA_Ensemble. This framework integrates a novel biologically inspired optimization algorithm, the Indian Millipede Optimization Algorithm (IMOA), for effective feature selection. It also incorporates an enhanced Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN-GP), which uses attention layers, layer normalization, and skip connections in the discriminator to improve the realism of synthetic minority-class samples. A dynamic attention-based ensemble (DA_Ensemble) comprising Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Feedforward Neural Network (FNN) models is employed to boost classification performance. The proposed model was evaluated on benchmark datasets including UNSW-NB15, CIC-IDS2017, and H23Q under both binary and multiclass settings. It achieved up to 99% accuracy, precision, recall, and F1-score. The results, validated through five-fold cross-validation and ablation studies, demonstrate consistent performance improvements across diverse scenarios.