Practical Network Coding with Resilient Subspace Codes

Network coding allows nodes in a network to combine different packets using linear operations. In most instances, the encoding coefficients are chosen randomly and placed in the packet header. The ability to correct errors and erasures is critical, because a single malicious packet injected by a misbehaving node or the deletion of a single packet can corrupt multiple packets and jeopardize the entire information flow. Building on Rotter and Kschischang´s theoretical work on subspace network codes, we propose a practical network coding protocol with in-built resilience against faults and active attacks. Also included is a low-complexity extended code construction for high-rate subspace Reed-Solomon like codes. The code maintains the distance properties that are key for error and erasure correction. Performance results show that throughput gains can be achieved with lower complexity and smaller field sizes.