Because the amount of spam mail has been increasing and it currently occupies 85% of all mail traffic, SMTP servers are frustrated with the task of filtering. To address this problem, it has been proposed that the SMTP server blocks spam mail before receiving the message body at the SMTP session by using host information or behaviors of spam sender clients. Although these proposals block spam mail at a high rate, they may also block legitimate mail, which causes a great deal of trouble to legitimate SMTP clients because they are forced to resend mail. We propose an efficient spam mail-blocking system based on information obtained before receiving the DATA command at the SMTP session. This method suppresses the false positive rate minimally and works as a lightweight primary filter. The preliminary result of our method was that it could detect  95% of spam mail.