Evaluating Trust in Ad Hoc Network Routing by Induction of Decision Trees

Abstract

Routing protocols play a vital role to ensure communication on the network, but by default do not address the issues of misbehavior. It is necessary to add mechanisms that make a better selection of paths for mitigating malicious nodes and increasing network performance. The dynamic nature of ad hoc networks requires distributed and adaptive mechanisms, allowing the removal and reintegration of nodes as they change their behavior and acting more severe or mild depending on the circumstances. In this paper we propose an evaluation mechanism that aims to mitigate routing misbehavior and other network failures. Four attributes of the routes are considered: level of activity, trust, mobility and number of hops. We performed simulations using the DSR protocol in scenarios with selfish nodes and dynamic topology in order to cause anomalies in the network. The routing information obtained containing the paths taken, the success rates and the attributes of the routes. To learn the best routes, we trained the decision tree by induction using the C 4.5 algorithm. The results show that the evaluation mechanism adapts to each context and creates harder rules on hostile and unstable environments. This mechanism could improve decision-making from routing protocols, providing greater trust to the network.