Unknown Flow Detection with Imbalanced Traffic Data: Poster Abstract

Abstract

Unknown traffic is a serious problem in network traffic classification for network management and security, since it degrades the performance of a traffic classifier. We propose a machine learning-based classifier that can not only categorize the traffic of known classes but also detect unknown traffic in data imbalanced network environment, where frequency of traffic from one application outweighs that of other application. We improved the previous RTC model by adding data preprocessing module and simulated with synthesized dataset to show the effect of the data imbalance problem on classification accuracy and to validate the performance of our model in data-imbalanced network environment.