FSF: Applying Machine Learning Techniques to Data Forwarding in Socially Selfish Opportunistic Networks

Abstract

Opportunistic networks are becoming a solution to provide communication support in areas with overloaded cellular networks, and in scenarios where a fixed infrastructure is not available, as in remote and developing regions. A critical issue, which still requires a satisfactory solution, is the design of an efficient data delivery solution trading off delivery efficiency, delay, and cost. To tackle this problem, most researchers have used either the network state or node mobility as a forwarding criterion. Solutions based on social behaviour have recently been considered as a promising alternative. Following the philosophy from this new category of protocols, in this work, we present our “FriendShip and Acquaintanceship Forwarding” (FSF) protocol, a routing protocol that makes its routing decisions considering the social ties between the nodes and both the selfishness and the device resources levels of the candidate node for message relaying. When a contact opportunity arises, FSF first classifies the social ties between the message destination and the candidate to relay. Then, by using logistic functions, FSF assesses the relay node selfishness to consider those cases in which the relay node is socially selfish. To consider those cases in which the relay node does not accept receipt of the message because its device has resource constraints at that moment, FSF looks at the resource levels of the relay node. By using the ONE simulator to carry out trace-driven simulation experiments, we find that, when accounting for selfishness on routing decisions, our FSF algorithm outperforms previously proposed schemes, by increasing the delivery ratio up to 20%, with the additional advantage of introducing a lower number of forwarding events. We also find that the chosen buffer management algorithm can become a critical element to improve network performance in scenarios with selfish nodes.