SpamAssassin also includes plugins to support reporting spam messages automatically or manually to collaborative filtering databases such as Pyzor, DCC, and VipulвЂ™s Razor. Specifically, SpamAssassin has been shown to produce around 1.5% false negatives (spam that was missed) and around 0.06% false positives (ham incorrectly marked as spam). I haven't had a complaint on that domain regarding spam in nearly 8 years.SpamAssassin typically differentiates successfully between spam and non-spam in between 95% and 100% of cases, depending on what kind of mail is received and training of its Bayesian filter. Since I use tagging instead of quarantining, my users don't even know it's there until they actually look into the headers of emails that are in their Junk Email folder. With the model I use, with Mailcleaner & Exchange virtualized and a spare Mailcleaner virtual-appliance image backed-up, I have very, very little maintenance or interaction with either system, except when I add a new user, and even then all I have to do is add the user to AD, normally, because Mailcleaner is using AD on the back-end to validate all email addresses & users. Personally, I use Active Directory (LDAP) auth, which is easy to setup if you're fluent with LDAP query structure, or at least have setup a Barracuda to talk to AD/LDAP, since it's basically the same sort of syntax. On it's own, with no modifications, it fully supports LDAP, SMTP, or local-authentication. Now, all of that said, Mailcleaner is very dependable, if installed according to the instructions. Otherwise, be sure to keep a spare copy of the VM ready to deploy just in case it goes belly-up. just be sure you know what you're doing when you get in there, if you do any mods. this is due to the custom structure that the developer(s) use, which doesn't correspond to Debian defaults (though Mailcleaner, itself, is based on Debian-Sarge). Mailcleaner can be very ticky under the hood. This ensures that if you should decide to get under the hood and do any experimentation or development on Mailcleaner, that you'll still have a fully working copy, ready to deploy at any time, that will only take about 10 to 20 minutes to return to service after a catastrophic failure of the VM due to altering. I've also used Mailcleaner on Physical hardware, and it worked great there too, but in the long run I've found that it's much better to virtualize Mailcleaner, shutdown the Mailcleaner VM and replicate it to a virtual appliance after you've configured Mailcleaner for your domain and authentication system(s), then bring Mailcleaner back online.

I have both Mailcleaner & Exchange running from within VMWare ESXi virtual servers, on the same host even, and I don't see any remarkable performance hit to the host or the VMs from doing so.

This ensures that my Mailcleaner anti-spam appliance never uses much disc-space, and ensures that my Exchange MessageStore doesn't outgrow the disc space allotted to it's virtual-machine-container. Personally, I don't use quarantining on because it's much less long-term maintenance to simply use tagging within MailCleaner and then let my Exchange server process tagged emails into the user's Junk Email folder, which has a 30 day expiry. I also like the fact that Mailcleaner supports tagging as well as quarantine.

While the Barracudas at my work see a lot more traffic overall, the percentage-rates-by-volume are nearly identical to what I experience with Mailcleaner. I've used Mailcleaner for several years now to protect my domain (), while using Barracuda M600s at work (). Mailcleaner is by far the best opensource spam filtering solution I've found, and I've tested pretty much any that I've been able to locate (mostly those integrated into other firewall-type-linux-distributions). I've gotta agree with John Gardeniers here.

Spamassassin alternatives