活动时间：2025-05-30 10:00
活动地点：2号学院楼2432
主讲人：徐茂超
主讲人中文简介：
徐茂超博士于2010年在美国波特兰州立大学获得统计学博士学位，现为伊利诺伊州立大学数学系终身教授，同时担任英国网络风险评估公司Rankiteo人工智能部门 Head。近年来主要开展基于统计方法的网络安全风险评估和治理研究，相关研究成果发表于Annals of Applied Statistics, Technometrics，IEEE Transactions on Information Forensics and Security，IISE Transactions 等国际著名期刊。研究曾获伊利诺伊州立大学杰出研究奖和Janice Witherspoon Neuleib 研究奖，北美精算师协会最佳论文奖，以及应用统计杂志最佳论文推荐奖等。徐教授还曾担任美国网络安全公司CouldCover，美国网络风险评估公司Safe Security以及新加坡虚拟货币保险公司InsurAce的学术顾问。
活动内容摘要：
Modeling the multivariate dependence among cyber data breach risks presents a significant challenge due to the sparsity and heavy tail properties exhibited by breach events. In this talk, we introduce a novel ensemble learning approach that effectively captures both the temporal and cross-sectional dependence inherent in cyber risks. Our approach leverages bivariate copulas to generate predictive members, and the resulting predictive distribution is carefully calibrated by minimizing the distribution score. Moreover, we demonstrate the applicability of our proposed model in the domain of insurance pricing. Through extensive simulations and analysis of real-world data, our findings reveal that our approach outperforms existing methodologies reported in the literature. The superior performance of our approach highlights its potential to enhance risk assessment and insurance pricing practices related to cyber data breaches.
主持人：孙春友