A Skype ML Datasets Validation and Detection Mechanism Using Machine Learning Approach

Internet traffic classification is an area of current research interest. Identification of real time applications such as Skype has gained more attention in the last few years. Skype traffic classification is challenging because Skype uses encrypted traffic and uses no well-known port number. Several methods which used both signature-based and statistical approaches were proposed. However, the training and testing datasets validation have not been formally addressed. This work highlights the problem of machine learning (ML) datasets validation and proposes a mechanism based on ML statistical approach to identify Skype traffic. Two different networks environment are considered for Skype traffic to gain insight into the statistical features of Skype traffic. Ten algorithms within Weka are used to examine the best algorithm for the given datasets. Random Forest was found to be the best resulting in more than 99.8% accuracy.