Deep Packet: A Novel Approach For Encrypted Traffic Classification Using\n Deep Learning

Internet traffic classification has become more important with rapid growth\nof current Internet network and online applications. There have been numerous\nstudies on this topic which have led to many different approaches. Most of\nthese approaches use predefined features extracted by an expert in order to\nclassify network traffic. In contrast, in this study, we propose a \\emph{deep\nlearning} based approach which integrates both feature extraction and\nclassification phases into one system. Our proposed scheme, called "Deep\nPacket," can handle both \\emph{traffic characterization} in which the network\ntraffic is categorized into major classes (\\eg, FTP and P2P) and application\nidentification in which end-user applications (\\eg, BitTorrent and Skype)\nidentification is desired. Contrary to most of the current methods, Deep Packet\ncan identify encrypted traffic and also distinguishes between VPN and non-VPN\nnetwork traffic. After an initial pre-processing phase on data, packets are fed\ninto Deep Packet framework that embeds stacked autoencoder and convolution\nneural network in order to classify network traffic. Deep packet with CNN as\nits classification model achieved recall of $0.98$ in application\nidentification task and $0.94$ in traffic categorization task. To the best of\nour knowledge, Deep Packet outperforms all of the proposed classification\nmethods on UNB ISCX VPN-nonVPN dataset.\n