Routing for Security in Networks with Adversarial Nodes

TL;DR

This paper develops LP-based routing schemes for secure unicast in networks with adversarial nodes, achieving rate-optimality among routing-only methods and outperforming naive network coding in certain cases.

Contribution

It introduces a novel LP-based routing scheme for secure communication against node adversaries, proving its rate-optimality among routing-only schemes and comparing favorably to network coding approaches.

Findings

LP-based routing scheme is rate-optimal among routing-only methods

Routing schemes outperform naive network coding in some scenarios

Extended routing schemes to node-jamming scenarios with similar results

Abstract

We consider the problem of secure unicast transmission between two nodes in a directed graph, where an adversary eavesdrops/jams a subset of nodes. This adversarial setting is in contrast to traditional ones where the adversary controls a subset of links. In particular, we study, in the main, the class of routing-only schemes (as opposed to those allowing coding inside the network). Routing-only schemes usually have low implementation complexity, yet a characterization of the rates achievable by such schemes was open prior to this work. We first propose an LP based solution for secure communication against eavesdropping, and show that it is information-theoretically rate-optimal among all routing-only schemes. The idea behind our design is to balance information flow in the network so that no subset of nodes observe "too much" information. Interestingly, we show that the rates achieved by our routing-only scheme are always at least as good as, and sometimes better, than those achieved by "na\"ive" network coding schemes (i.e. the rate-optimal scheme designed for the traditional scenario where the adversary controls links in a network rather than nodes.) We also demonstrate non-trivial network coding schemes that achieve rates at least as high as (and again sometimes better than) those achieved by our routing schemes, but leave open the question of characterizing the optimal rate-region of the problem under all possible coding schemes. We then extend these routing-only schemes to the adversarial node-jamming scenarios and show similar results. During the journey of our investigation, we also develop a new technique that has the potential to derive non-trivial bounds for general secure-communication schemes.