GAN-based imbalanced data intrusion detection system

TLDR: The purpose of this study is to solve data imbalance by using the Generative Adversarial Networks (GAN) model, which is an unsupervised learning method of deep learning which generates new virtual data similar to the existing data.

Abstract: According to the development of deep learning technologies, a wide variety of research is being performed to detect intrusion data by using vast amounts of data. Although deep learning performs more accurately than machine learning algorithms when learning large amounts of data, the performance declines significantly in the case of learning from imbalanced data. And, while there are many studies on imbalanced data, most have weaknesses that can result in data loss or overfitting. The purpose of this study is to solve data imbalance by using the Generative Adversarial Networks (GAN) model, which is an unsupervised learning method of deep learning which generates new virtual data similar to the existing data. It also proposed a model that would be classified as Random Forest to identify detection performance after addressing data imbalances based on a GAN. The results of the experiment showed that the performance of the model proposed in this paper was better than the model classified without addressing the imbalance of data. In addition, it was found that the performance of the model proposed in this paper was excellent when compared with other models that were previously used widely for the data imbalance problem.