Cyber-risk Management Framework for Online Gaming Firms: an Artificial Neural Network Approach

Abstract

Hackers have used Distributed-Denial-of-Service attacks to overwhelm a firm’s cyber-resources resulting in disrupted access to legitimate end-users. Globally, DDoS attacks cost firms between US$ 120 K to US$ 2 M for each incident. Apart from the monetary loss, they also disrupt service quality and damage the brand reputation of firms. In 2018-2019, Massively Multiplayer Online Gaming (MMOG) firms witnessed 74% of the total DDoS attacks. MMOG firms form a lucrative segment for hackers because of their large customer base and the massive incentive to cause disruptions and losses. Our Feedforward Neural Network-based Cyber-risk Assessment and Mitigation (FNN-CRAM) model consists of three modules: assessment, quantification, and mitigation. The cyber-risk assessment module uses FNN, which takes seven inputs comprising DDoS attack intensity and duration for five DDoS attack types, vulnerability data (i.e., their counts and score), and the vulnerability trends over time. This layer is connected to a ten-neuron hidden layer and one neuron output layer that estimates the probability of these attacks. We also observe that the probability of these DDoS attacks follows a Weibull distribution. Next, our cyber-risk quantification module computes the expected loss. We note that expected losses due to these DDoS attacks follow a gamma distribution. Our cyber-risk mitigation module uses a heat matrix to help the CTO (i) prioritize the cyber-risk associated with a DDoS attack and (ii) decide whether to reduce, accept, or pass the cyber-risk using technological and cyber-insurance interventions.