Abstract: Security is the key issue for broadly applying mobile agent technology to distributed computing environment. The promise based security model is theoretically proposed towards preventing the malicious behaviors of agents in the interaction process from the point of view of probability. In the environment with ability limited detecting mechanism, it is hard to determine the illegibility of a malicious agent, so there is the need for extra mechanism preventing the malicious behaviors of agents. Promise based mechanism means that in the agent interaction process, an agent is required to provide some amount of assurance cost so that the malicious behaviors of agents can be restrained. It can be proved that the rational assurance cost is feasible.