In this paper, we introduce a novel approach to correct packet errors and packet losses in store and forward by using binary error-correcting codes. Using a framework similar to what has been proposed for error control in random linear network coding, we investigate error control under two scenarios. First, we show that the Hamming metric is suitable for error control in the case of errors intrinsic to the network. Second, we investigate the case of an adversary on the network who erases and injects packets in order to corrupt the communication. Under this setting, we show that error correction is performed using a new metric, referred to as the modified Hamming metric. We then investigate using constant-weight codes and linear codes for error correction in store and forward. We thus show that the traditional approach of indexing the packets in order to recover their original order is a suboptimal restriction of our approach.
展开▼
