Machine Learning Assisted Cyber Loss Estimation Tool for CISOs
Open Access DepositedSegment Based Frequency Model
An important challenge throughout cybersecurity governance is cyber riskquantification. Chief Information Security Officers are often challenged with defending budget requests from the C-suite, selecting insurance coverage needed, and communicating risk clearly including in financial terms. However, the frameworks they use today do not have the financial precision needed nor do they estimate costs by company size or specific segments for decision support. This praxis develops a machine learning enhanced compound Poisson framework that provides segment level annual cyber loss distributions in order to address this gap CISOs face today. The pipeline built in the study decomposes cyber loss into frequency and severitycomponents. Both these components are modeled independently and then combined through Monte Carlo simulation. Histogram Gradient Boosting model configured with Poisson loss function is used in the frequency portion of the framework. Poisson loss function has the ability to handle sparse, skewed count data without requiring distributional assumptions making it the right fit for frequency modeling. On the other hand, the severity portion uses empirical bootstrap sampling. When combined these components provide full loss distributions across 63 industry-size segments. From this Value at Risk and Tail Value at Risk metrics are derived to support insurance structuring and capital allocation decisions. The final results are presented using an interactive dashboard for CISOs to use.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.