Trust is used in many applications, such as on-line auctions. However, trust establishment is not well understood. Different researchers have focused on different aspects of trust establishment. In this paper, we provide a unifying view of trust establishment, which includes the primitives, authentication, semantics of trust and trust evaluation. We propose a model which can serve as a framework for trust establishment. Trust is justified only if its establishment is secure against an adversary who seeks to gain it under false pretenses. We define an adversary model that takes malicious activities into account. We have conducted experiments to demonstrate how different policies react to attacks using an on-line auction as the scenario.