A Platform for Expressive and Secure Data Sharing with Untrusted Third Parties

[摘要] Today, third-party applications provide a variety of rich services to smartphone users. There are compelling reasons to share personal data with third parties, such as social networking applications, but the benefits of sharing data must be balanced against the corresponding risks to personal privacy. Prior work has proposed personal data vaults to separate the capturing and sharing of data. We argue that to express realistic access control policies, personal data vaults must support the ability to flexibly transform data before release (e.g., converting location data to the zip code level). We propose a simple, practical personal data vault design that supports full-fledged computation by both trusted and untrusted entities. Our design includes a small scripting language called TRANSMUTE which pipes data streams through sandboxed filters written in the Lua programming language. A straightforward analysis of TRANSMUTE scripts provides a strong guarantee of noninterference, and our approach admits a flexible mechanism for audit that provides finer-grained information about data leakage. We formalize our approach and prove a noninterference theorem. We also describe our implementation and evaluate its expressiveness and performance through several case studies.

[发布机构] UCLA Henry Samueli School of Engineering and Applied Science

[学科分类] 计算机科学（综合）