On Capacity Region of Wiretap Networks

In this paper we consider the problem of secure network coding where an adversary has access to an unknown subset of links chosen from a known collection of links subsets. We study the capacity region of such networks, commonly called "wiretap networks", subject to weak and strong secrecy constraints, and consider both zero-error and asymptotically zero-error communication. We prove that in general discrete memoryless networks modeled by discrete memoryless channels, the capacity region subject to strong secrecy requirement and the capacity region subject to weak secrecy requirement are equal. In particular, this result shows that requiring strong secrecy in a wiretap network with asymptotically zero probability of error does not shrink the capacity region compared to the case of weak secrecy requirement. We also derive inner and outer bounds on the network coding capacity region of wiretap networks subject to weak secrecy constraint, for both zero probability of error and asymptotically zero probability of error, in terms of the entropic region.