Application protocols identification using Non-parametric Estimation method

Application protocols identification (APPI) is of fundamental importance to many network tasks. Traditional payload-based methods are difficult to identify new encrypted protocols. Many machine learning methods cannot achieve high performance. In this paper, we address an identification method based on Bayesian theory and Non-parametric Estimation which has good performance in the field of pattern recognition. The new method firstly abstracts the statistical features from the payloads of the flows. Then application protocols are identified with Non-parametric Estimation method. Experimental results show that the new method achieves over 80% in F-score for protocols identification.