Date Added: Nov 2009
The authors investigate the secure capacity of arbitrary networks in the presence of Byzantine adversary. In their previous work and they show by example that linear network coding cannot achieve secure network capacity for error detection. To the best of their knowledge, it was the first work that identified the insufficiency of linear network codes in achieving secure capacity, even for unicast. In this paper, they extend this idea to error correction. They introduce the example of the Diamond Network, for which the error correction capacity can be arbitrarily larger than the maximum rate linear codes can achieve, and the error detection capacity is twice as large as the maximum rate linear codes can achieve.