Spam mails distributed from botnets have been one of the critical problems for the Internet. Spamming is growing at a rapid rate since sending a flood of mails is easy and very cheap. Spam mails waste user time and consume resources e.g., space and network bandwidth, so fighting against spam is an interesting issue in computer security. We have spent for more than 3 years collecting and analyzing over 161,230 emails from several mailboxes. We found that some users received up to 235 emails per day, only 1 to 3 emails were legitimate and the rest appeared to be spam mails. This paper presents a fast effective spam filter by analyzing the mail header. It works well with both text-base spam and all kinds of image spam. Our experiments and results showed that spam was filtered out at least 96.23% with no false positive.