Cyber risk : an analysis of self-protection and the prediction of claims

• Main Author: Azevedo, Alana K.
• Publication Date: 2021
• Other Authors: Bergel, Agnieszka I., Reis, Alfredo D. Egídio dos
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10400.5/24492
• topic: Cyber risk; Cybersecurity; Propensity Score; Neural Network; Multilayer Perceptron

For a set of Brazilian companies, we study the occurrence of cyber risk claims by analyzing the impact of self protection and the prediction of their occurrence. We bring a new perspective to the study of cyber risk analyzing the probabilities of acquiring protection against this type of risk by using propensity scores. We consider the problem of whether acquiring cyber protection improves network security using a matching method that allows a fair comparison among companies with similar characteristics. Our analysis, assisted with Brazilian data, shows that despite informal arguments that favor self-protection against cyber risks as a tool to improve network security, we observed that in the presence of self-protection against cyber risks, the incidence of claims is higher than if there were no protection. Regarding the prediction of the occurrence of a claim, a system considering a feedforward multilayer perceptron neural network was created, and its performance was measured. Our results show that, when applied to the relevant information of the companies under study, it presents a very good performance, reaching an eciency in general classication above 85%. The fact is that the use of neural networks can be quite opportune to help in solving the problem presented.