Classification of BGP Anomalies Using GRU with BGP Update Messages

With the increasing reliance on the Internet, its reliability and security have become major concerns. The Border Gateway Protocol (BGP) is susceptible to anomalies such as hijacking, configuration errors, and denial-of-service attacks, which can pose significant threats to the performance and reliability of the Internet. However, with the methods for the BGP anomalies classification, the operators can understand the cause of anomalies and take action to solve the problems. Recently, various techniques have been proposed for classifying BGP anomalies utilizing machine learning models. Nevertheless, we have identified some limitations of these classification models that raise doubts regarding their applicability in real-world scenarios for classifying new anomalies. In order to better classify BGP anomalies with high accuracy and efficiency, we introduce Gate Recurrent Unit (GRU) to classify BGP anomalies based on the features selected from BGP update messages. Notably, our method boasts a leaner parameter set and converges more rapidly compared to alternative approaches. Experimental results demonstrate that our method outperforms other methods, with the performance of the precision, recall and F1-score are 100%, 77% and 87% respectively.