Coverage analysis of information dissemination in throwbox-augmented DTN

TL;DR

This paper models information dissemination in throwbox-augmented DTNs using bipartite network theory, deriving a simple equation to predict coverage and revealing coverage invariance over time.

Contribution

It introduces a bipartite network model to analyze DTN coverage and derives a closed-form equation for accurate coverage prediction, surpassing traditional epidemic models.

Findings

Coverage is limited by fixed popular places regardless of time.

Coverage can be predicted using the largest component in the bipartite network.

Increasing places makes full coverage more difficult, while activity variation improves coverage.

Abstract

This paper uses a bipartite network model to calculate the coverage achieved by a delay-tolerant information dissemination algorithm in a specialized setting. The specialized Delay Tolerant Network (DTN) system comprises static message buffers or throwboxes kept in popular places besides the mobile agents hopping from one place to another. We identify that an information dissemination technique that exploits the throwbox infrastructure can cover only a fixed number of popular places irrespective of the time spent. We notice that such DTN system has a natural bipartite network correspondence where two sets are the popular places and people visiting those places. This helps leveraging the theories of evolving bipartite networks (BNW) to provide an appropriate explanation of the observed temporal invariance of information coverage over the DTN. In this work, we first show that information coverage can be estimated by the size of the largest component in the thresholded one-mode projection of BNW. Next, we also show that exploiting a special property of BNW, the size of the largest component size can be calculated from the degree distribution. Combining these two, we derive a closed form simple equation to accurately predict the amount of information coverage in DTN which is almost impossible to achieve using the traditional techniques such as epidemic or Markov modeling. The equation shows that full coverage quickly becomes extremely difficult to achieve if the number of places increases while variation in agent's activity helps in covering more places.