A MACHINE LEARNING APPROACH TO CYBER BREACH PREDICTION USING SUPPORT VECTOR MACHINES

Abstract

The evolution and development of cyber dangers can only be understood with the use of data collected from cyber occurrences. Since this field of study is still in its infancy, there is much more to discover. Cyberespionage activities, primarily malware infections, that occurred between 2005 and 2017 are the focus of this statistical study. We found that distributions based on autocorrelation do not adequately describe the frequency and severity of cyber attacks. Rather, random procedures work better. Then, we display the bespoke stochastic process models we developed to account for different breach magnitude and inter-breach time span. On top of that, we prove that these models are highly predictive of both arrival timeframes and damage levels. We use quantitative and qualitative trend studies to find out how cyber events have evolved over the years. Concerning security, one can say many things. The frequency of leaks has increased, but the total number of leaks has remained unchanged.