Performance evaluation of spam filter is an important process for choosing and improvement of spam filter. Nowadays, the evaluation does not based on realistic environment. Users do not participate in the process and privacy problems are always encountered. In this research, we propose a framework which evaluate in realistic environment. The framework uses a corpus of real e-mails. Users are volunteers who evaluate their e-mails without privacy problems. This is an ease of use and cheap framework. It uses a sampling method for reducing volunteers’ jobs. Volunteers in this framework are an ideal spam filter which always gives correct classification. The result show a capability of this framework for spam filters evaluation and volunteer have quick response to our process