Enhanced Tab Transformer with Month-Based Binning Missing Data Imputation for Cyber Security Situation Prediction
Zihan Xiong, Jun Chen, Dabei Chen, Yuan Feng, Xuesong Guo · Advances in transdisciplinary engineering · 2024
Network security situation prediction is a critical task aimed at analyzing existing threats in the current network environment and predicting potential attacks in the future. This paper proposes an Enhanced Tab Transformer with a Month-Based Binning Missing Data Imputation method to effectively address the NSSP task. However, the original network data presents challenges, such as missing data noise and difficult data forms for training. To address this issue, we use an ensemble machine learning method based on month-based binning to fill a large amount of missing data. Additionally, we improve the tab transformer to handle NSSP’s tabular data form and predict network security status. Our proposed method is evaluated on the publicly available CNCERT dataset. Experimental results demonstrate that our proposed method has good classification performance compared to existing techniques.