It is well known that trust relationships between entities are uncertain in pervasive computing. However, previous researchers did not attach enough importance to uncertainty in their trust models. Therefore, their models always seem arbitrary when they set some specific values to various trust levels. To remedy this problem, we declare that uncertainty should be taken into account while modeling trust. This paper presents such an approach, in which we model trust based on cloud theory, which is a promising theory to describe uncertain concepts. In other words, we regard trust between entities as a cloud, say trust cloud. Based on such a definition of trust, we further propose algorithms to compute propogated trust clouds and aggregated trust clouds, which are needed in case of trust reasoning for entities in pervasive computing environments. Besides, we also propose an approach to use our cloud-based trust modelin pervasive applications, which is helpful to put our model into practice in trust-based security mechanisms.