SaTC: CORE: Small: A Privacy-Preserving Meta-Data Analysis Framework for Cyber Abuse Research - Foundations, Tools and Algorithms

该项目开发并部署了一个移动通信元数据分析平台，旨在预测青少年骚扰或虐待事件，利用参与式方法推动研究设计决策并维护青少年隐私。通过与青年参与者合作分析传播模式（而不是内容），该项目通过促进对青少年虐待和骚扰的情境化理解做出了创新贡献。反过来，这一分析支持开发新的增强隐私的工具，以便在虐待和痛苦的情况下进行早期干预。从被虐待或痛苦的青少年表现出明显不同的交流模式的假设出发，收集的数据使研究人员能够以新的方式可视化和分析青少年的日常交流网络，为了解青少年社会性和虐待形式的结构和条件提供了新的机会，同时开发了新的预测模型和技术。特别是，这个项目1)设计了基于参与式设计的新方法，可以更好地理解年轻人将隐私、成人监督和在线监控概念化的方式；2)以终端用户的隐私需求为核心设计原则，开发可扩展性高、可实际部署的移动平台，进行大规模元数据和调查分析；3)确保平台保证参与者的完全匿名性，同时还允许纵向和灵活的调查管理，元数据收集和无缝扩展；4)从通信元数据日志（从文本和呼叫）中提取与网络骚扰和虐待相关的时空新特征；5)利用提取的特征设计上下文感知机器学习算法，从元数据日志中快速检测异常模式。该项目培养了一批从事跨学科研究的研究生和本科生。该项目的长期影响远远超出了青年人口的范围，该项目更普遍地朝着一个完全匿名、保密、基于需求的社会科学研究平台的发展。

This project develops and deploys a mobile communications metadata analysis platform, designed to predict incidents of youth harassment or abuse, drawing on participatory methods to drive research design decisions and maintain youth privacy. By analyzing communication patterns - rather than content - in collaboration with youth participants, this project makes innovative contributions by facilitating a contextualized understanding of youth abuse and harassment. In turn, this analysis supports the development of new, privacy-enhancing tools which allow for early intervention in instances of abuse and distress. Starting from the hypothesis that abused or distressed youth display observably different patterns of communication, the data collected allow researchers to visualize and analyze the everyday communication networks of youth in new ways, providing new opportunities to understand the structures and conditions of youth sociality and forms of abuse while simultaneously developing new predictive models and techniques.In particular, this project 1) designs new methods grounded in participatory design that allows for a better understanding of the ways in which youth conceptualize privacy, adult supervision and surveillance online; 2) develops a highly scalable and practically deployable mobile platform for performing large scale metadata and survey analysis, taking the privacy needs of end users as a core design principle as identified from the first outcome; 3) ensures that the platform guarantees full anonymity to participants, while also allowing for longitudinal and flexible survey administration, metadata collection, and seamless scaling; 4) extracts novel features in the spatial and temporal domain from communication metadata logs (from texts and calls) that associate with cyber harassment and abuse; 5) leverages the features extracted to design context aware machine learning algorithms to quickly detect abnormal patterns from metadata logs. The project trains a number of graduate and undergraduate student in inter-disciplinary research. Long term impact of this project scale well beyond youth populations, and the project more generally moves towards the development of a fully anonymous, confidential, needs-based social scientific research platform.