Enabling Private Real-Time Applications by Exploiting the Links Between Erasure Coding and Secret Sharing Mechanisms

Abstract

A huge amount of personal data is shared in real time by online users, increasingly using mobile devices and (unreliable) wireless channels. There is a large industry effort in aggregation and analysis of this data to provide personalised services, and a corresponding research effort to enable processing of such data in a secure and privacy preserving way. Secret sharing is a mechanism that allows private data sharing, revealing the information only to a select group. A parallel research effort has been invested in addressing the performance of real time mobile communication on lossy wireless channel, commonly improved by using erasure codes. In this thesis, we bring together the theoretically related fields of secret sharing and erasure coding, to provide a rich source of solutions to the two problem areas. Our aim is to enable solutions that deliver the required performance level while being efficient and implementable. The thesis has the following contributions.

We evaluate the applicability of a new class of Maximum Distance Separable (MDS) erasure codes to transmission of real time content to mobile devices and demonstrate that the systematic code outperforms the non-systematic variant in regards to computation complexity and buffer size requirements, making it practical for mobile devices.

We propose a new Layered secret sharing scheme for real time data sharing in Online Social Networks (OSNs). The proposed scheme enables automated profile sharing in OSN groups with fine-grained privacy control, via a multi-secret sharing scheme comprising of layered shares. The scheme does not require reliance on a trusted third party. Compared to independent sharing of specific profile attributes (e.g. text, images or video), the scheme does not leak any information about what is shared, including the number of attributes and it introduces a relatively small computation and communications overhead.

Finally, we investigate the links between MDS codes and secret sharing schemes, motivated by the inefficiency of the commonly used Shamir scheme. We derive the theoretical links between MDS codes and secret sharing schemes and propose a novel MDS code based construction method for strong ramp schemes. This allows the use of existing efficient implementations of MDS codes for secret sharing and secure computing applications. We demonstrate that strong ramp schemes deliver a significant reduction of processing time and communication overhead, compared to Shamir scheme.