SaTC: CORE: Medium: Collaborative: Enabling Long-Term Security and Privacy through Retrospective Data Management

在线数据存储，从过去的对话到纳税申报表到游戏约会邀请，都可以完全忠实地保留数年或数十年。虽然保存在在线档案中的数据不会改变，但围绕它们的个人和社会背景会改变。这些生活变化可能需要更改或删除存储的数据，但不幸的是，用户在线档案中的大量数据使手动管理变得不可行。该项目的目标是开发方法和工具，使用户能够管理他们多年来积累的数据，利用以用户为中心的设计和机器学习来部分自动化该过程。这些工具将使人们能够更好地了解在现代长期在线档案的背景下的追溯隐私。它们还将使用户能够更有效地管理这些档案中的风险。与研究界分享的发现将推动发现超越这个项目。随着时间的推移，随着上下文的变化，对用户安全和隐私概念化的理解一直受到该领域内缺乏广泛，仔细收集的数据集的阻碍。该项目将在用户许可的情况下收集匿名数据集，以便在这一领域开展进一步研究。该团队正在进行首批纵向研究之一，研究所需的安全和隐私决策如何随时间变化。该项目还将收集关于用户对长期数据的风险和效用的看法以及追溯管理机制的可接受性的定性见解。此外，目前还不了解时间性如何影响机器学习技术用于隐私，也不了解如何捕捉概念漂移以确保可以在巨大的档案中识别潜在的威胁。这些任务需要新的机器学习方法和预测模型，这些方法和模型既可以考虑时间维度，又可以在自动化归档管理时最大限度地减少用户负担。最后，该项目将设计和实施新颖的以用户为中心的界面，以满足目前尚未满足的需求，帮助用户有效地最大限度地减少其大型长期在线档案中的安全和隐私风险。该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估来支持。

Online data storage, everything from past conversations to tax returns to playdate invitations, may be retained at full fidelity for years or decades. Although the data being saved in online archives does not change, the personal and social contexts surrounding them do. Those life changes may necessitate changing or deleting stored data but, unfortunately, the vast quantity of data in users' online archives makes manual management infeasible. The goal of this project is to develop methods and tools that enable users to manage the data they have accumulated over many years, leveraging user-centered design and machine learning to partially automate the process. These tools will enable a better understanding of retrospective privacy in the context of modern long-lived online archives. They will also empower users to more effectively manage the risks embedded in these archives. The findings shared with the research community will advance discovery beyond this project. The understanding of user conceptualizations of security and privacy over time, as contexts change, has been stymied by a lack of broad, carefully collected datasets within this domain. This project will collect anonymized datasets, with users' permission, that enable further research in this area. The team is conducting one of the first longitudinal studies of how desired security and privacy decisions change over time. The project will also gather qualitative insights about users' perceptions of risk and utility for long-term data, as well as the acceptability of retrospective management mechanisms. Furthermore, it is not currently understood how temporality impacts the use of machine learning techniques for privacy, nor how to capture concept drift to ensure that latent threats can be identified within immense archives. These tasks require new machine learning approaches and predictive models that can both account for the temporal dimension and minimize user burden when automating archive management. Finally, the project will design and implement novel user-centered interfaces that address the currently unmet need of helping users efficiently minimize security and privacy risks in their large, long-term online archives.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.