Using Deep Learning and Graph Analysis against Cyberattacks

If an IT network is modelled as an abstract graph with network devices represented as nodes in the graph, and connections between the components represented as edges, you can already derive interesting information from the resulting topological structure. If in addition the network traffic on this graph is captured, machine learning algorithms can be used to identify anomalous behaviour caused by network intruders.In this paper we'll look at the fundamentals of graph analytics and how these technologies can be used to detect anomalies in general. We will show how graph analysis can be combined with machine learning using the integration between Oracle Big Data Spatial and Graph and R through the OAAgraph package. And finally, we will describe a project in which network data was analysed by means of a deep learning engine to detect suspicious network activity.