A0765
Title: Estimation and accuracy evaluation of cyber-risk prioritization for threat intelligence
Authors: Mario Angelelli - University of Salento (Italy) [presenting]
Serena Arima - University of Salento (Italy)
Christian Catalano - University of Salento (Italy)
Enrico Ciavolino - University of Salento (Italy)
Abstract: The pervasive diffusion of interconnected Information and Communication Technologies (ICTs) is driving the need to properly assess cyber risk and prioritize counteractions, aiming to better manage resources for cybersecurity in the prevention of cyber incidents. The increasing complexity of digital systems requires flexible models for extracting the information necessary to ensure data integrity, confidentiality, and availability. Motivated by this need, a new statistical methodology is introduced to support cyber-vulnerability prioritization expressed in terms of ordinal data, which assesses the severity of a vulnerability. The new method combines a non-parametric regression model and a new accuracy index meeting operative requirements often encountered in the cybersecurity field. Specifically, the methodology uses mid-quantile regression as a robust approach for ordinal severity assessments, and the proposed accuracy measure enjoys invariance properties for consistent ranking derivation. The proposed model is tested on simulated and real data, which are obtained from the fusion of different databases providing relevant information on exploiting cyber vulnerabilities. The proposal is compared with alternative methods (ordered logit, linear regression for rank-transformed variables) to evaluate the potential advantages of this approach, the domains where these advantages are significant, and their interpretation for threat intelligence.
