Secure random linear network coding on a wiretap network

We develop a secure random linear network coding scheme on wiretap networks where a wiretapper can only eavesdrop on a limited number of channels. On one hand, by refining Lima's “locked coefficients” method and applying the approach of one-time pad, our scheme can well protect message packets without decreasing network throughput. On the other hand, by treating ciphertext as noisy symbols, inspired by the physical layer technique, and applying Shamir's secret sharing scheme, our scheme can successfully protect secret random seed without any forms of key exchange or secret channels. Compared to existing schemes, our scheme has minimum information overhead, independency of hash functions, and no restriction on global encoding kernel. Finally, we analyze the computational complexity of our proposed scheme and rigorously prove that our scheme can achieve secure network communication.