Encrypted internet traffic classification using a supervised spiking neural network

Internet traffic recognition is an essential tool for access providers since recognizing categories related to different data packets transmitted on a network help them define adapted priorities. That means, instance, high priority requirements audio conference and low ones file transfer, enhance user experience. As internet becomes increasingly encrypted, the mainstream classic technique, payload inspection, rendered ineffective. This paper uses machine learning techniques encrypted classification, looking only at packet size time of arrival. Spiking neural networks (SNN), largely inspired by how biological neurons operate, were used two reasons. Firstly, they are able recognize time-related features. Secondly, can be implemented efficiently neuromorphic hardware with energy footprint. Here we very simple feedforward SNN, one fully-connected hidden layer, trained in supervised manner using newly introduced method known as Surrogate Gradient Learning. Surprisingly, such SNN reached accuracy 95.9% ISCX datasets, outperforming previous approaches. Besides better accuracy, there also significant improvement simplicity: input size, number neurons, trainable parameters all reduced four orders magnitude. Next, analyzed reasons this good accuracy. It turns out that, beyond spatial (i.e. size) features, exploits temporal ones, mostly nearly synchronous (within 200ms range) arrival times certain sizes. Taken together, these results show that SNNs excellent fit classification: more accurate than conventional artificial (ANN), could power embedded systems.