Cloud Environment Intrusion Detection Based on GAN and LightGBM

In the context of rapid advancements in cloud computing technology, a significant number of cloud service providers are facing the impact of extensive malicious attacks, necessitating the urgent development and implementation of intrusion detection systems (IDS). Researchers worldwide have proposed numerous IDS solutions to counter the threat of network intrusions. However, the performance of intrusion detection systems noticeably deteriorates when dealing with unbalanced data. Although extensive research has been conducted on unbalanced data, most studies suffer from inherent weaknesses, such as potential data loss or overfitting. To address these challenges, this paper introduces a robust intrusion detection system based on the integration of generative adversarial network (GAN) and lightweight gradient boosting machine (LightGBM). Experimental results on the NSL-KDD and UNSW-NB15 datasets demonstrate the capability of the proposed model to improve the detection rate for a select subset of samples, outperforming traditional state-of-the-art models in terms of accuracy, recall, and F-score values.