Available online at www.sciencedirect.com 1877–0509 © 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of Prof. Mitsuhisa Sato and Prof. Satoshi Matsuoka doi:10.1016/j.procs.2011.04.190 Procedia Computer Science 4 (2011) 1751–1760 International Conference on Computational Science, ICCS 2011 A Data-Driven Framework for Dynamic Trust Management Olufunmilola Onolaja ∗ , Georgios Theodoropoulos, Rami Bahsoon School of Computer Science, The University of Birmingham, United Kingdom, B15 2TT Abstract Reputation and trust-based models have been used extensively in different application domains. These include large online communities such as eBay, Amazon, YouTube and ad-hoc and wireless sensor networks. Recently, the use of the models has gained popularity due to their effectiveness in providing trusted systems or networks. These models focus on online and historical data to determine the reputation of domain members. In this paper, we propose a novel approach for obtaining trust values by focusing not only on online and historical data but also possible future scenarios to anticipate events in the next time intervals. The data-driven framework is able to dynamically obtain and inject data to predict the future trust value of every identity in the system. The advantage of this proactive approach compared to other approaches is that informed decisions about the domain can be made before a compromise occurs. Keywords: Reputation, Trust, Data-driven systems 1. Introduction Reputation and Trust-based models (RTMs) have gained popularity over the years, borrowing ideas from game theory and Bayesian networks. RTMs are described as systems that provide mechanisms to produce a metric encap- sulating reputation for each identity in a given application domains [1]. Generally, RTMs aim to provide information that allow nodes to distinguish between trustworthy and untrustworthy members. The models encourage members to cooperate through the use of incentives, and discourage maliciousness by punishment schemes such as isolation and service denial. RTMs have been adopted in applications that rely on the cooperation of domain members in order for the appli- cation to function correctly. The models have been used extensively in various e-commerce and online communities such as YouTube, Amazon and eBay as described in Section 2. Some literatures also suggest their use in domains ranging from peer-to-peer (P2P) to mobile networks [2, 3, 4]. A common problem of RTMs is their vulnerability to collusion attacks, where two or more nodes can team up to behave maliciously. Incentive policies that are used in P2P networks to ensure cooperation between nodes are generally susceptible to collusion attacks as well. Traditionally, the models rely on recommendations that are based on past interactions between the members provided by the same members to decide on the reputation of one another. ∗ Corresponding author Email addresses: O.O.Onolaja@cs.bham.ac.uk (Olufunmilola Onolaja), G.K.Theodoropoulos@cs.bham.ac.uk (Georgios Theodoropoulos), R.Bahsoon@cs.bham.ac.uk (Rami Bahsoon)