Abstract:
Cybersecurity risk modeling is a relatively new topic that has attracted the attention of companies seeking to provide insurance coverage against cyberattacks. In this study I introduce Dynamic Generalized Poisson panel-data models for cybersecurity risk modeling. Following Zhu(2012) I extend the results of the Generalized Poisson INGARCH Model to the case of panel data with partial and complete pooling. As an application, I use cyberattack data on 491 consecutive victim IP addresses which exhibit intrinsic spatiotemporal attack patterns (as analyzed by Chen et.al (2015)). After the models are estimated I compare them according to their likelihood value, AIC and BIC criteria. Finally, I provide a forecast comparison of some of the models. The results of this study can be further used in the cyber-insurance industry for example, for the pricing of insurance products.
