Andrade Cesar, Bragança Hendrio, Feitosa Eduardo, Souto Eduardo
Institute of Computing, Federal University of Amazonas, Amazonas, Brazil.
Data Brief. 2023 Dec 21;52:109999. doi: 10.1016/j.dib.2023.109999. eCollection 2024 Feb.
In the pursuit of advancing research in continuous user authentication, we introduce COUNT-OS-I and COUNT-OS-II, two distinct performance counter datasets from Windows operating systems, crafted to bolster research in continuous user authentication. Encompassing data from 63 computers and users, the datasets offer rich, real-world insights for developing and evaluating authentication models. COUNT-OS-I spans 26 users in an IT department, capturing 159 attributes across diverse hardware and software environments over 26 h on average per user. COUNT-OS-II, on the other hand, encompasses 37 users with identical system configurations, recording 218 attributes per sample over a 48-hour period. Both datasets utilize pseudonymization to safeguard user identities while maintaining data integrity and statistical accuracy. The well-balanced nature of the data, confirmed by comprehensive statistical analysis, positions these datasets as reliable benchmarks for the continuous user authentication domain. Through their release, we aim to empower the development of robust, real-world applicable authentication models, contributing to enhanced system security and user trust.
在推进持续用户认证研究的过程中，我们引入了COUNT-OS-I和COUNT-OS-II，这是两个来自Windows操作系统的不同性能计数器数据集，旨在促进持续用户认证方面的研究。这些数据集涵盖了63台计算机和用户的数据，为开发和评估认证模型提供了丰富的真实世界见解。COUNT-OS-I涵盖了一个IT部门的26名用户，平均每位用户在26小时内跨越不同硬件和软件环境捕获了159个属性。另一方面，COUNT-OS-II涵盖了37名具有相同系统配置的用户，在48小时内每个样本记录218个属性。两个数据集都使用假名化来保护用户身份，同时保持数据完整性和统计准确性。经全面统计分析证实，数据的均衡特性使这些数据集成为持续用户认证领域可靠的基准。通过发布这些数据集，我们旨在推动强大的、适用于现实世界的认证模型的开发，为增强系统安全性和用户信任做出贡献。
