Stationary mobile behavioral biometrics: A survey
Stationary mobile behavioral biometrics: A survey
复制标题
DOI：
10.1016/j.cose.2023.103184
复制
发表时间：
2023-03
期刊：
影响因子：
--
通讯作者：
Aratrika Ray-Dowling;Daqing Hou;S. Schuckers
中科院分区：
文献类型：
--
作者：
Aratrika Ray-Dowling;Daqing Hou;S. Schuckers
当前移动设备中的安全机制（例如 PIN、密码、图案化密码和生物识别技术）是一次性入口点身份验证，容易受到攻击。此外，多重身份验证 (MFA) 等高级机制也会给用户体验带来摩擦。相比之下，行为生物识别技术依赖于用户与计算设备的交互来对用户进行身份验证，因此可以是连续的、非侵入性的且具有成本效益的，代表了补充现有身份验证技术的有前途的方向。该调查重点关注通过加速度、陀螺仪、磁力计和方向（旋转）等运动事件进行的静止/非行走（坐、站）移动行为生物识别，并可选择其他非运动、零星模式（如滑动和击键）的支持。对固定行为的关注是合理的，因为此类行为代表了用户与移动设备交互的主要方式。为了帮助读者了解用户活动/行为的广阔前景，我们将最先进的行为分为自然行为和设计行为，并描述认知心理学中行为生物识别的基础。此外，我们根据运动模式的融合将调查研究分为三组，并按照任务、数据集、模式、算法和性能等维度描述每项研究的特征。根据我们的调查，我们确定了几个未来的研究方向。
Current security mechanisms in mobile devices such as PINs, passwords, patterned passwords, and biometrics are one-time entry-point authentication and vulnerable to attacks. Furthermore, advanced mechanisms like Multi-Factor Authentication (MFA) introduce friction in the user experience. In contrast, behavioral biometrics rely on user interaction with computing devices to authenticate a user and thus, can be continuous, non-intrusive, and cost-effective, representing a promising direction that complements existing authentication techniques. This survey focuses on stationary/non-walking (sitting, standing) mobile behavioral biometrics through motion events like acceleration, gyroscope, magnetometer, and orientation (rotation) with the optional support of other non-motion, sporadic modalities such as swipes and keystrokes. The focus on stationary behaviors can be justified because such behaviors represent the major way a user interacts with mobile devices. To help readers understand the broad landscape of user activities/behaviors, we categorize the state of the art intonaturalanddesignedbehaviors and describe the underpinning of behavioral biometrics in cognitive psychology. Furthermore, we categorize the surveyed studies into three groups based on the fusion of motion modalities and characterize each study along dimensions such as task, datasets, modality, algorithms, and performance. Based on our survey, we identify several future directions of research.