A Feasibility Study of Watchdogs on Opportunistic Mobile Networks
Diogo Soares, Bruno Matthaus, Edjair Mota, Celso B. Carvalho · 2018
Detection of selfish behavior is crucial to the proper functioning of opportunistic mobile networks operations. Previous works focused on developing reputation systems or trust models. However, these approaches assume the existence of an efficient model for detecting selfish nodes, such as watchdogs. Moreover, other studies do not present an experimental analysis of the efficiency of the watchdog implementation. This article discusses the design, implementation, and validation of a watchdog, in order to verify the effectiveness of the detection process and to understand the effects of this mechanism for increasing density of messages in the network. We tested the proposed watchdog with two different mobility traces, with a low and a high contact density. The analysis of true positives present a very promising result for watchdog evaluation on real opportunistic mobile networks. However, our results showed that the implementation of watchdog may present some issues related to how false positives were detected during the monitoring and in case that the number of contacts increases. To handle this, we propose an acknowledgement mechanism to increase the reliability of false positive detections and the results improved above 68% in both traces. Our results also present a brief study of congestion control and, even with a congestion control of six hours, the results remain virtually unchanged, so that staying with forwarding information for a long time does not affect reliability. The improvement in the number of messages relayed was above 35% for both traces.