Tuesday, February 28, 2012
IST Lunch Bunch
Error Correction for Deadlines, Keys and Computationally Limited Networks
Tracey Ho, Assistant Professor, Electrical Engineering and Computer Science, Caltech
Classical coding theory deals with the problem of robust transmission on individual unreliable channels. Its network generalization considers coding for networks where some links may be erroneous. Network error correction coding for single-source multicast networks has been well studied. We show that problems with multiple sources or multiple receivers with different demands require new coding schemes and capacity bounding techniques, and introduce a variety of applications. One application is streaming communication, where specific information must be decoded by specific deadlines. This can be modeled as a network error/erasure correction problem where each deadline corresponds to a receiver with different demands. We use this perspective to characterize capacity and design good codes. In decentralized networks, our results on secure multi-source network error correction can be used for distributing cryptographic keys when not all participating nodes are trustworthy. Finally, we construct rateless network error correction codes allowing coding-based information theoretic security, which uses path diversity and redundant capacity as resources, to supplement cryptographic security, which relies on computation as a resource, in networks of computationally limited low-power devices.