Enhancing intrusion detection in IoT networks: CWInfoGAN based data augmentation approach

Intrusion detection in Internet of Things (IoT) networks faces significant challenges due to highly imbalanced datasets and the continuous emergence of evolving cyberattacks. Minority attack classes are often underrepresented, leading to biased learning and reduced detection performance. To address this issue, we develop CWInfoGAN, an extension of InfoGAN that prioritizes the generation of minority class samples. Unlike standard InfoGAN, which generates all attack types equally, this may lead to biased learning and detection of underrepresented attacks. Experimental results on the TON-IoT dataset indicate that CWInfoGAN-based augmentation provides competitive classification performance while enabling controlled generation of underrepresented attack classes. The findings are specific to the evaluated TON-IoT setting, and further validation on additional IoT intrusion-detection datasets is required to assess generalizability.