S. Fischer-Hübner et al. (Eds.): TrustBus 2006, LNCS 4083, pp. 182 – 191, 2006. © Springer-Verlag Berlin Heidelberg 2006 Trust Model Architecture: Defining Prejudice by Learning M. Wojcik, J.H.P. Eloff, and H.S. Venter Information and Computer Security Architectures Research Group (ICSA) Department of Computer Science, University of Pretoria {hibiki}@tuks.co.za, {eloff, hventer}@cs.up.ac.za Abstract. Due to technological change, businesses have become information driven, wanting to use information in order to improve business function. This perspective change has flooded the economy with information and left businesses with the problem of finding information that is accurate, relevant and trustworthy. Further risk exists when a business is required to share information in order to gain new information. Trust models allow technology to assist by allowing agents to make trust decisions about other agents without direct human intervention. Information is only shared and trusted if the other agent is trusted. To prevent a trust model from having to analyse every interaction it comes across – thereby potentially flooding the network with communications and taking up processing power – prejudice filters filter out unwanted communications before such analysis is required. This paper, through literary study, explores how this is achieved and how various prejudice filters can be implemented in conjunction with one another. 1 Introduction Technological development has influenced the principles required to run a successful economy [1]. However, the advent of new technologies and the subsequent implementations thereof have resulted in exposure to new risks. Two risk factors exist that continually drive research towards lessening the risks encountered: effective communication and security. In order to accomplish an organisation’s desired task, effective and timely communication is required. An organisation makes use of technology to communicate and share information. This information is an asset to the organisation and is used to assist decision-making processes. It is important that this information be reliable and accurate so that it can be trusted [2]. Trust models have been proposed in order to minimise the risk of sharing and successfully analysing information [3], [4]. Trust models rely on the abstract principle of trust in order to control what information is shared and with whom. Trust models evaluate the participants of a transaction and assign a numerical value, known as a trust value, to the interaction. This numerical value is used to determine the restrictions placed on the transaction and the nature of information shared. This process of analysis