Mobile agents are software entities consisting of code and data that can migrate autonomously from host to host executing their code. Despite its benefits, security issues strongly restrict the use of code mobility. The protection of mobile agents against the attacks of malicious hosts is considered the most difficult security problem to solve in mobile agent systems. In [3] the authors introduced the idea of limiting the execution time in the hosts. Malicious hosts need time to analyze and modify an agent in order to take some profit. Controlling the execution time in the hosts permits detecting manipulation attacks performed by malicious hosts during the agent?s execution. This paper present a protocol for detecting malicious hosts based on the idea of execution time limiting.