DNSBL (RBL) Comparison


In an attempt to find a good, low false positive RBL, I analysed my mail server logs writing a couple of scripts. I have put up this page so that other mail server administrators have some figures before them while choosing their list of RBLs.
I do not advise use of more than three (remote) RBLs since it usually involves serious performance penalty.

Methodology


Testing was done for minimum of 10 days of each month. Each day a cron job parsed my qmail-smtpd log files and got a list of unique IPs, which were checked by a perl script against the said RBLs. As a side effect of the exercise, I also checked whether the IPs had their reverse dns (rdns) set up since I gather from experience that refusing to accept mails from IPs having no rdns blocks a huge number of spam.

Note: I recommend you all to have a look at UCEPROTECT-Blacklists. Their false positive rate is not very high, their technical support is responsive but their main USP for me is that they allow their zone files to be downloaded free of charge unlike other RBLS like Spamhaus or Spamcop.

For ease of comparison, figures given below are expressed as percentage of total unique IPs.

Time Frame zen.spamhaus.org dnsbl-1.uceprotect.net bl.spamcop.net b.barracudacentral.org dyna.spamrats.com Missing RDNS
Mar 09 81.4 62.8 67.3 80.3 -- 33.8
Apr 09 90.1 72.2 75.2 88.5 42.6 34.2
May 09 90.2 72.7 75.2 88.9 43.8 32.6
Jun 09 90.3 73.4 75.4 87.6 37.0 34.9
Jul 09 93.3 77.9 81.4 89.5 38.7 39.8
Aug 09 92.6 81.1 82.1 92.3 39.0 39.7
Sep 09 94.1 73.2 72.1 94.2 43.3 27.2
Oct 09 95.6 79.4 74.4 95.3 38.1 27.9
Dec 09 89.9 82.3 60.4 90.1 37.3 26.1
Jan 10 85.4 81.1 50.3 78.9 32.3 20.5