A Survey of Traffic Classification Validation and Ground Truth Collection

Traffic classification has recently been widely used in quality of service measurements, network management and security monitoring. With the development of internet, many traffic classification methods have been proposed and developed. Correctly validate performance of traffic classifiers requires a perfect ground truth, i.e., information about the application or the protocol of traffic flows. However, the validation of the existing classification approaches is a controversial problem, because a perfect ground truth is difficult to obtain. Recently, several researchers have proposed solutions to collect ground truth of traffic traces and achieved satisfactory results. This paper presents review on the state of the art of traffic classification validation and ground truth collection, and explores the challenges and future directions.