Secure Quasi-Realtime Collaborative Editing over Low-Cost Storage Services

Abstract

Abstract.A realtime collaborative editor facilitates concurrent editing of a document by multiple authors. It is desired that the document be shared only among the authors, and protected from the potentially curious server. Existing approaches have taken two distinct paths -centralized server based approaches that achieve high concurrency and meet real-time requirement but compromise on security and incur high server cost, and peer-to-peer based approaches that support security but compromise on users' convenience and mobility. In this paper, we observe that by relaxing the realtime requirement, we can achieve security, reduce server cost and yet exploit the conveniences of the centralized setting. In particular, we consider generic low-cost storage servers in the cloud that provide storage integrity but do not guarantee low-latency. Essentially, our method breaks the document into small encrypted regions that are stored on the server and coordinates the authors' access. Although two authors are unable to concurrently modify a same region, the system is able to provide "quasi-realtime" experience. By relaxing the requirement to quasi-realtime, the difficulties in achieving document consistency, and the requirement on resources are significantly reduced. We give a proof-of-concept implementation on top of Dropbox, a commercial cloud storage service. Preliminary user studies show that the system is effective.

Spacetime Characterization of Real-Time Collaborative Editing

“…Fourth, our results give insights on how to design end-to-end encrypted RTCE tools, where users need not to trust platform hosters for content confidentiality [8,34]. A key decision in designing such tools is whether block-or stream-ciphers should be used.…”

“…Considering that most of the collaboration happens in either space or time, but not in spacetime, our findings suggest that section-based conflict avoidance, as implemented for instance by MediaWiki, would be enough to cater for author needs. Intermediate synchronization granularities have been partially explored in the area of cryptographically secure pads [8,34], but there are no final answers yet. Our experiments show that this design space for RTCE should be explored further to find a sweet spot between complexity and adhering to author needs.…”