This paper is concerned with trust modeling for networked computing systems. Of particular interest to this paper is the observation that trust is a subjective notion that is invisible, implicit and uncertain in nature, therefore it may be suitable for being expressed by subjective probabilities and then modeled on the basis of Bayesian principle. In spite of a few attempts to model trust in the Bayesian paradigm, the field lacks a global comprehensive overview of Bayesian methods and their theoretical connections to other alternatives. This paper presents a study to fill in this gap. It provides a comprehensive review and analysis of the literature, showing that a large deal of existing work, whether or not proposed based on Bayesian principle, can cast into a general Bayesian paradigm termed subjective Bayesian trust (SBT) theory here. The SBT framework can thus act as a general theoretical infrastructure for comparing or analyzing theoretical ties among existing trust models, and for developing novel models. The aim of this study is twofold. One is to gain insights about Bayesian philosophy in modeling trust. The other is to drive current research step ahead in seeking a high-level, abstract way of modeling and evaluating trust. A Survey on Trust Modeling from a Bayesian Perspective