A Bayesian Generalized Poisson Model for Cyber Risk Analysis

Cyber threats are now considered as a top risk for many economic sectors which include retail, financial services, security, and healthcare. The costs for damages from cyber-attacks and the number of cyber-attacks are two of the main quantities of interest when measuring cyber-risk. In this paper, we focus on the frequency of cyber-attacks and analyse some features through the lens of a generalized Poisson model. We follow a Bayesian inference approach and apply a Markov Chain Monte Carlo algorithm for posterior approximation. In the application to a well-known dataset on cyber-threats we find evidence of over-dispersion and of time-variations in the features of the phenomenon.