Distributed Network Behaviour Prediction Using Machine Learning and Agent-Based Micro Simulation

Abstract

Distributed network behaviour is increasingly attracting huge attention both in academics and industrial initiatives, and most recently machine learning has been used in every possible field, leveraging its advantages.Typically, a distributed networking allows for the execution of distributed applications which results in complex behaviours among connected systems. This complexity in itself can create grey areas and vulnerabilities in the securities of these networks, therefore predicting the behaviour of these systems both at the macro and micro level has become essential. In the past most researchers have predicted network behaviour and network attack patterns by using aggregated data, but in this paper we focus on the application of machine learning at the individual user level such that the prediction of the individual network user behaviour pattern at the micro level becomes a substasive tool in creating a realistic agent based simulation of the whole distributed network, which in turn can serve as a test bed for predicting what-if scenarios such as network attacks on the target system or exposing vulnerabilities within the target system. The simulation result was validated by comparing the simulated interaction within the simulated network to the data logged in the server logs within the real life network system. This produced a correlation above 0.8, indicating a realistic model.