Perfectly Secure Communication, based on Graph-Topological Addressing in Unique-Neighborhood Networks
This addresses secure communication in networks for applications requiring perfect security, but it appears incremental as it builds on existing graph-topological concepts without demonstrated broad impact.
The paper tackles the problem of achieving perfectly secure communication without end-to-end shared secrets or computational assumptions by introducing unique-neighborhood networks, where nodes are uniquely identifiable based on their graph-topological neighborhoods, enabling techniques for confidentiality, authenticity, and availability.
We consider network graphs $G=(V,E)$ in which adjacent nodes share common secrets. In this setting, certain techniques for perfect end-to-end security (in the sense of confidentiality, authenticity (implying integrity) and availability, i.e., CIA+) can be made applicable without end-to-end shared secrets and without computational intractability assumptions. To this end, we introduce and study the concept of a unique-neighborhood network, in which nodes are uniquely identifiable upon their graph-topological neighborhood. While the concept is motivated by authentication, it may enjoy wider applicability as being a technology-agnostic (yet topology aware) form of addressing nodes in a network.