BEACON (Behavioral Engine for Authentication & Continuous Monitoring)
收藏资源简介:
BEACON是由塔帕尔工程技术学院与滑铁卢大学联合创建的大规模多模态数据集,旨在通过高保真电竞环境捕获细粒度行为指纹,以推动连续身份验证与行为分析研究。该数据集包含约430GB同步模态数据,总计461GB磁盘存储,涵盖28名玩家在102.51小时《Valorant》竞技游戏中产生的超过9000万次鼠标事件、49.8万次击键事件及1.14亿个网络数据包,数据来源包括高频率硬件输入、网络抓包及屏幕录像等多维度信息。数据集通过定制化低延迟日志架构在真实游戏过程中采集,严格保障了多模态数据的时间同步性与环境上下文完整性。其核心应用领域为网络安全,特别专注于解决高认知负荷场景下连续身份验证系统的鲁棒性评估、用户行为画像构建以及多模态表征学习等关键问题。
BEACON is a large-scale multimodal dataset jointly created by Thapar Institute of Engineering and Technology and the University of Waterloo. It aims to capture fine-grained behavioral fingerprints in a high-fidelity esports environment to advance research on continuous authentication and behavioral analysis. This dataset contains approximately 430GB of synchronized multimodal data, with a total disk storage footprint of 461GB. It covers over 90 million mouse events, 498,000 keystroke events, and 114 million network packets generated by 28 players during 102.51 hours of competitive Valorant gameplay. The data is collected from multi-dimensional sources including high-frequency hardware input, network packet capture, and screen recording. The dataset is collected during actual gameplay using a customized low-latency logging architecture, which strictly ensures the temporal synchronization of multimodal data and the integrity of environmental context. Its core application domain is cybersecurity, with a particular focus on addressing key issues such as robustness evaluation of continuous authentication systems, user behavior profiling, and multimodal representation learning in high cognitive load scenarios.

- 1BEACON: A Multimodal Dataset for Learning Behavioral Fingerprints from Gameplay Data塔帕尔工程技术学院·计算机科学与工程系; 滑铁卢大学 · 2026年



