Personal domain mail where this machine is the mail exchange goes through a number of counter-spam techniques listed below. More detailed discussions likely occured on these matters on W3C Team-only mailing lists before they were carried out. No techniques are adopted without careful consideration of risks to detering legitimate mail. The techniques being adopted are some of the ones being evaluated or also adopted for W3C's own mail. Spammers constantly evolve to try to avoid their mail from being intercepted and rejected. We will continue to explore counter techniques as they become available. Not all counter measures are listed below.
Our goal is to reduce spam received by anyone who gets mail through this machine, including forwarding addresses. Should you experience any problems with these measures please submit a detailed report, including bounced mail less any personal message within it to whoever owns the domain for which your mail had problems. That W3C Team member can contact the system administrators should the matter not be something they can handle themselves.
- Exim The MTA we chose has a number of sanity checks it does which help reduce junk mail from the start. For instance it only accepts mail claiming to be from legitimate domains by confirming dns records exist.
- SPF This emerging standard provides a means for domains to specify where mail from that domain can originate from. If you have a domain that receives mail on this machine and all the people in that domain consistently send mail using certain mailservers you should collect the IP addresses of them and create spf record inside your dns records to reduce forgeries from your domain.
- Spamassassin This is antispam software that runs on all inbound mail through this server. It performs a number of heuristic tests maintained by a group of developers and also does bayesian training on spam/ham. Spamassassin uses a scoring system based on all the tests applied. Mail that scores 10 or above is rejected. All mail gets a header added containing the score it is X-Spam-MTA-Score. Mail with a score of 4 or more is very likely spam, there is the risk of false positives which is why we set the score higher for rejecting. You may wish to filter your mail with procmail or in your mail client based on the score in this header.
- Vipul's Razor This is an external database of known spam based on signature of message bodies. It is tied into Spamassassin.
- Blacklists. W3C maintains blacklists of domains and individuals who have clearly demonstrated themselves as spam offenders. People.w3.org keeps it's blacklists up to date with these blacklists leveraging them. There is no subscription to external blacklists since some of them have proven from our experience to be erroneous and block legitimate mail.
- ClamAV This is an antivirus engine that updates local databases of known viruses off of central, maintained database on the net. Mails identified as containing viruses are bounced.
- Honey Pots. We have a number of email addresses that have been put out to get onto spammers lists that should receive no legitimate mail. Any mail to these addresses goes to a honeypot address and used to train our Spamassassin bayes database as spam to improve the likelihood it will trap and discard future spam of a similar nature.
Ted Guild <ted@w3.org>
