Epidemic-like Proximity-based Traffic Offloading

TL;DR

This paper presents a theoretical framework and algorithm for selecting seed users in proximity-based communication to maximize cellular traffic offloading in mobile social networks, achieving over 60% offloading with few seeds.

Contribution

It introduces a gossip-style social cascade model and a greedy algorithm for seed selection, providing near-optimal solutions for traffic offloading in epidemic-like information diffusion.

Findings

Proximity-based communication can offload over 60% of cellular traffic.

The greedy algorithm achieves about 63% of the optimal traffic offloading.

The approach outperforms heuristic and random seed selection methods.

Abstract

Cellular networks are overloaded due to the mobile traffic surge, and mobile social network (MSNets) carrying information flow can help reduce cellular traffic load. If geographically-nearby users directly adopt WiFi or Bluetooth technology (i.e., leveraging proximity-based communication) for information spreading in MSNets, a portion of mobile traffic can be offloaded from cellular networks. For many delay-tolerant applications, it is beneficial for traffic offloading to pick some seed users as information sources, which help further spread the information to others in an epidemic-like manner using proximity-based communication. In this paper, we develop a theoretical framework to study the issue of choosing only k seed users so as to maximize the mobile traffic offloaded from cellular networks via proximity-based communication. We introduce a gossip-style social cascade (GSC) model to model the information diffusion process, which captures the epidemic-like nature of proximity-based communication and characterizes users' social participation as well. For static networks as a special-case study and mobile networks, we establish an equivalent view and a temporal mapping of the information diffusion process, respectively, leveraging virtual coupon collectors. We further prove the submodularity in the information diffusion and propose a greedy algorithm to choose the seed users for proximity-based traffic offloading, yielding a solution within about 63% of the optimal value to the traffic offloading maximization (TOM) problem. Experiments are carried out to study the offloading performance of our approach, illustrating that proximity-based communication can offload cellular traffic by over 60% with a small number of seed users and the greedy algorithm significantly outperforms the heuristic and random algorithms.