Enabling Retrospective Management of Data in the Cloud

Online cloud storage is a convenient and affordable way to store data over long periods. Today, millions of Internet users who have adopted cloud storage have accumulated years of information across thousands of files. The data stored is diverse, ranging from old photo-albums to old tax returns. As the social and personal contexts around this data continue to evolve, some of it loses its value and relevance over time. Even worse, some may contain sensitive information that puts users at risk. With the increasing prevalence of cyber crimes and data breaches, the decision to retain certain information online should be re-evaluated over time. Unfortunately, due to the scale of data, manual management of the cloud is infeasible, and there is a need for smart tools that allow users to achieve their desired management in an effective manner. In this thesis, we present a comprehensive oversight into the problem. Through an exploratory study, we investigate the need for remediations for retrospective data management. Our results demonstrate a clear desire among users to manage files and the inability to do so due to the lack of practical tools. We then conduct a second study in which we carry out qualitative interviews to understand the kinds of management users intend. Finally, we incorporate the learned insights into the design of a learning-based tool (Aletheia). We predict users desired management decisions with an accuracy of 79%. Aletheia’s performance validates a human-centric approach to developing management for files in the cloud. It also improves upon state of the art in minimizing the attack surface of cloud accounts.