A novel algorithm for encrypted traffic classification based on sliding window of flow's first N packets

Network applications are getting more and more prevalent along with the development and the widespread use of encrypted network applications. However, traffic classification methods may need to be improved to realize more stable classification in a more sufficient way. Here, we proposed a novel Sliding Window First N Packets algorithm for the encrypted network traffic classification. With this method, one could evidently reduce the flow characteristics feature dimension, as well as the number of packets in each traffic flow. The experimental results show that under a reduced dimension of encrypted traffic flow characteristics and also a reduced number of each flow data packets, average classification accuracy using the Sliding Window First N Packets algorithm we proposed is more than 95%. By using our approach, one can achieve a general increase of the traffic classification accuracy by about 3% compared with the existing methods.