Intensive experiences show and confirm that grid environments can be considered as the most promising way to solve several kinds of problems relating either to cooperative work especially where involved collaborators are dispersed geographically or to some very greedy applications which require enough power of computing or/and storage. Such environments can be classified into two categories; first, dedicated grids where the federated computers are solely devoted to a specific work through its end. Second, Volunteer grids where federated computers are not completely devoted to a specific work but instead they can be randomly and intermittently used, at the same time, for any other purpose or they can be connected or disconnected at will by their owners without any prior notification. Each category of grids includes surely several advantages and disadvantages; nevertheless, we think that volunteer grids are very promising and more convenient especially to build a general multipurpose distributed scalable environment. Unfortunately, the big challenge of such environments is, however, security and trust. Indeed, owing to the fact that every federated computer in such an environment can randomly be used at the same time by several users or can be disconnected suddenly, several security problems will automatically arise. In this paper, we propose a novel solution based on identity federation, agent technology and the dynamic enforcement of access control policies that lead to the design and implementation of trusted volunteer grid environments.