Recently, various schemes against spam are proposed because of rapid increasing of spam. Some schemes are based on sender whitelisting with auto registration, a principle that a recipient reads only messages from senders who are registered by the recipient, and a sender have to perform some procedure to be registered (challenge-response.) In these schemes, some exceptions are required to show error mail to a sender of an original message. However, spammers can abuse this exception to send spam to users. We have proposed improved scheme in [1], combining challenge-response and Bayesian filtering. In this paper, we make tests on our scheme and a scheme using only Bayesian filtering to show efficiency of our scheme.