Abstract
Traffic classification, i.e. the identification of the type of applicationsflowing in a network, is a strategic task for numerous activities (e.g.,intrusion detection, routing). This task faces some critical challenges thatcurrent deep learning approaches do not address. The design of currentapproaches do not take into consideration the fact that networking hardware(e.g., routers) often runs with limited computational resources. Further, theydo not meet the need for faithful explainability highlighted by regulatorybodies. Finally, these traffic classifiers are evaluated on small datasetswhich fail to reflect the diversity of applications in real-world settings. Therefore, this paper introduces a Lightweight, Efficient andeXplainable-by-design convolutional neural network (LEXNet) for Internettraffic classification, which relies on a new residual block (for lightweightand efficiency purposes) and prototype layer (for explainability). Based on acommercial-grade dataset, our evaluation shows that LEXNet succeeds to maintainthe same accuracy as the best performing state-of-the-art neural network, whileproviding the additional features previously mentioned. Moreover, we illustratethe explainability feature of our approach, which stems from the communicationof detected application prototypes to the end-user, and we highlight thefaithfulness of LEXNet explanations through a comparison with post hoc methods.