Quinn Comendant wrote: > Thanks for the reply Michele. > > However, my question still applies even if the spam threshold > were increased. In some cases false positives are marked with > a score that is well beyond what I have found a effective threshold > for our networks. Here is an example: > >> X-Spam-Report: >> * 1.6 BAYES_60 BODY: Bayesian spam probability is 60 to 70% >> * [score: 0.6887] >> * 4.9 RCVD_IN_XBL RBL: Received via a relay in Spamhaus XBL >> * [67.125.228.229 listed in sbl-xbl.spamhaus.org] >> * 1.7 RCVD_IN_NJABL_DUL RBL: NJABL: dialup sender did non-local >> SMTP >> * [67.125.228.229 listed in combined.njabl.org] >> * 0.1 RCVD_IN_SORBS_DUL RBL: SORBS: sent directly from dynamic IP >> address >> * [67.125.228.229 listed in dnsbl.sorbs.net] >> * -1.1 AWL AWL: Auto-whitelist adjustment >> X-Spam-Status: Yes, hits=7.3 required=5.0 tests=AWL,BAYES_60, >> RCVD_IN_NJABL_DUL,RCVD_IN_SORBS_DUL,RCVD_IN_XBL autolearn=no > > The question again: how to make spamassassin more lenient for > dynamic RBLed IPs? Is the best solution simply to lower the score for > RBL rules? You may need to look at the RBLs that hit dynamic Ips and put in an over-ride on their scoring, so that they are pushed down. 4.9 for XBL seems very high! What are you getting for Bayes? We generally find that Bayes brings the potential FPs back into line with reality We use a much higher threshold than you do and block very effectively :)
Mr Michele Neylon Blacknight Internet Solutions Ltd Hosting, co-location & domains http://www.blacknight.ie/ Tel. +353 59 9137101 -- Email scanned by Blacknight for viruses and dangerous content. Visit http://www.blacknight.ie for more information
