Distributed Repair of Digital Broadcasts

Abstract

Digital broadcasts are a highly efficient way of transmitting data to a large number of receivers, but they are typically not designed for reliable transmission. In this thesis we attempt to design a fully peer-to-peer repair system that can operate on digital broadcasts that have no inherent support for reliability. We exchange hashed representations of the broadcasted stream between peers and use data differencing techniques along with majority voting to detect errors in the transmission with a high probability of success. We find that the theory behind the system is sound, but that transmitters in large broadcasting networks may slightly modify the transmitted data in such a way that extensive canonicalization is required for data differencing techniques to function properly. The presented system offers increased robustness compared to existing systems, but is significantly more complex