Behavioural Network Traffic Analytics for Securing 5G Networks

The analysis of the network traffic in 5G networks is of high significance to the network security administrator, since it could allow for the identification of different behavioural groups and the distinction of anomalous from normal activity. The problem is the multi-dimensional nature of the data, e.g. SMS, call, Internet, services etc. that makes it difficult to analyse. This is even more challenging in 5G networks, compared to previous generation networks, since one more dimension is added to the traffic, representing different network slices. In this respect, activity that is normal in one slice can be anomalous in another. This paper presents a graph-based method for network mining and visualization of user activities in a mobile network. The raw multi- dimensional network traffic data are used for the construction of multiple multi-dimensional graph- based features that capture specific behavioural aspects for each user. Within each feature, graph matching techniques are applied in order to identify groups of users with similar behaviour. The dissimilarity results for each feature are combined using a multi-objective visualization method. The outcome is a data visualization in which users with similar behaviour are depicted as points close to each other. The network analyst is able to select the desired trade-off among the multiple features, and visually detect groups of users with similar behaviours, as well as possible anomalous clusters or outliers. Experimental evaluation of the proposed approach in several application scenarios verify its efficiency.