A recommendation approach for user privacy preferences in the fitness domain
A recommendation approach for user privacy preferences in the fitness domain
复制标题
DOI：
10.1007/s11257-019-09246-3
复制
发表时间：
2019-10-12
影响因子：
3.6
通讯作者：
Knijnenburg, Bart P.
中科院分区：
文献类型：
--
作者：
Sanchez, Odnan Ref;Torre, Ilaria;Knijnenburg, Bart P.
健身追踪器无疑越来越受欢迎。随着这些设备持续捕获、存储和处理与健身相关的数据，确保用户隐私的需求变得越来越迫切。在本文中，我们应用数据驱动的方法来开发健身设备的隐私设置建议。我们首先提出了一个健身数据隐私模型，我们定义该模型以明确的方式表示用户的隐私偏好，符合欧盟的通用数据保护条例（GDPR），并能够表示用户和第三方的偏好。我们的众包数据集是使用健身领域的当前场景收集的，并通过应用机器学习技术用于识别隐私配置文件。然后，我们研究了不同的个人跟踪数据和用户特征，这些数据和特征可能会推动向用户推荐隐私配置文件。最后，在此基础上设计了一组具有不同指导风格的隐私设置推荐策略。有趣的是，我们的结果显示了用户特征、特征和态度之间的一些语义关系，这些关系对提供隐私建议很有用。尽管存在一些关于隐私偏好建模的工作，但本文在物联网和健身领域的数据共享和处理隐私偏好建模方面做出了贡献，并特别关注GDPR合规性。此外，识别识别良好的偏好集群和这些集群的预测因子对于用户分析和旨在平衡用户对其隐私权限的控制和设置这些权限的简单性的交互式推荐策略的设计有重要贡献。
Fitness trackers are undoubtedly gaining in popularity. As fitness-related data are persistently captured, stored, and processed by these devices, the need to ensure users' privacy is becoming increasingly urgent. In this paper, we apply a data-driven approach to the development of privacy-setting recommendations for fitness devices. We first present a fitness data privacy model that we defined to represent users' privacy preferences in a way that is unambiguous, compliant with the European Union's General Data Protection Regulation (GDPR), and able to represent both the user and the third party preferences. Our crowdsourced dataset is collected using current scenarios in the fitness domain and used to identify privacy profiles by applying machine learning techniques. We then examine different personal tracking data and user traits which can potentially drive the recommendation of privacy profiles to the users. Finally, a set of privacy-setting recommendation strategies with different guidance styles are designed based on the resulting profiles. Interestingly, our results show several semantic relationships among users' traits, characteristics, and attitudes that are useful in providing privacy recommendations. Even though several works exist on privacy preference modeling, this paper makes a contribution in modeling privacy preferences for data sharing and processing in the IoT and fitness domain, with specific attention to GDPR compliance. Moreover, the identification of well-identified clusters of preferences and predictors of such clusters is a relevant contribution for user profiling and for the design of interactive recommendation strategies that aim to balance users' control over their privacy permissions and the simplicity of setting these permissions.