GAN-Augmented Intrusion Detection: A Feature-Wise Attention MLP Framework for Imbalanced Network Traffic

Abstract

With the increase in Internet of Things applications, there is also an increase in refined cyberattacks. Earlier, there were a limited number of classes under cyberattacks, but now, there are numerous classes. Some of them are distinguishable, and some of them are very tough to be identified. Approaches involving convolutional neural networks, multilayer perceptrons (MLPs), and Long Short Term Memory (LSTM) have their own limitations in terms of generalization. Thus, this research represents a GAN-Augmented Attention-Enhanced MLP intrusion detection system, which we named as GAN-MLP. It uses Wasserstein Generative Adversarial Networks with Gradient Penalty to generate synthetic minority attacks that look real. The feature-wise attention mechanism of MLP helps to gain unequal and inequitable features and helps the model to outperform as compared to other models. The dataset used was CICIDS2017, and all the metrics were calculated on this dataset. The model achieved 96.5% accuracy, 97.12% precision, 96.71% F1-score, and recall of 96.5%. The Receiver-Operating Characteristics and Area Under the Curve (ROC-AUC) values consisting of micro-average 0.9992 and macro-average 0.9977 confirm the model’s remarkable performance. Thus, it is well suited for real-time intrusion detection and also for imbalanced network systems. The empirical analysis of the suggested methodology substantiates the efficacy of this strategy in comparison to the traditional deep learning approach.    Received: 19 June 2026 | Revised: 29 July 2026 | Accepted: 14 August 2026   Conflicts of Interest The authors declare that they have no conflicts of interest to this work.   Data Availability Statement Data sharing is not applicable to this article as no new data were created or analyzed in this study.    Author Contribution Statement Osha Shukla: Software, Validation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision. Annu Mishra: Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision. Mohd Dilshad Ansari: Conceptualization, Software, Validation, Resources, Data curation, Writing – review & editing, Visualization, Supervision, Project administration. Ravi Prakash Chaturvedi: Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration. Rajneesh Kumar Singh: Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Supervision.