A Combination-Based GAN for Enhanced Intrusion Detection in IoT

With the rapid expansion of the Internet of things (IoT), an increasing volume of sensitive information is being uploaded to online platforms. This growth has been accompanied by the emergence of complex and sophisticated intrusion behaviors. As a result, research into network-based intrusion detection has become critically important. This paper addresses the challenges of imbalanced and complex network traffic data in intrusion detection systems (IDS). A method that combines a Deep Convolutional Generative Adversarial Network (DCGAN) integrated with a self-attention mechanism and an independent classifier incorporating a Gated Recurrent Unit (GRU). The model employs dual-stage deep learning to optimize the discriminator's ability to identify attack samples effectively. Experimental results on the CSE-IDS-2018 dataset demonstrate that this approach performs well compared to various other intrusion detection techniques in terms of detection accuracy, recall, and F1-score for minority class attacks. This paper provides valuable insights into how combining GRUs and GANs can enhance the detection of malicious activities in cybersecurity contexts.