Am 18.02.2016 um 12:29 schrieb Matus UHLAR - fantomas:
Am 17.02.2016 um 17:49 schrieb Alex:http://pastebin.com/zKWUUQ0Q Obviously they're coming in advance of being on an RBL or DNSBL. I was thinking to correlate the body text somehow with something that checks to see if it actually passed through Google (SPF, etc?), but that won't work for messages that were forwarded to another user...On 17.02.16 20:17, Reindl Harald wrote:well, and that's why bayes-autoexpire is nonsense, your pastebin would have been rejected here by exceed 8.0 points (milter-rejcts score) easilybut only if you manually bump scores, which most of people should not. It also does in no way indicate that bayes autoexpire is nonsense, it only says it's better to use BAYES (and to have it properly cofigured)
such mails are coming over years in waves with weeks and month not appear and the tokes would expire as well as i recently faced other rejected spam trained a year ago and not seen for a long time
Content analysis details: (13.1 points, 5.5 required) pts rule name description ---- ---------------------- -------------------------------------------------- 7.5 BAYES_99 BODY: Bayes spam probability is 99 to 100% [score: 1.0000] 0.4 BAYES_999 BODY: Bayes spam probability is 99.9 to 100% [score: 1.0000] 2.5 HTML_IMAGE_ONLY_12 BODY: HTML: images with 800-1200 bytes of words 0.0 HTML_MESSAGE BODY: HTML included in message 2.5 RDNS_NONE Delivered to internal network by a host with no rDNS 0.2 HTML_SHORT_LINK_IMG_1 HTML is very short with a linked image 0.0 T_REMOTE_IMAGE Message contains an external imagescore HTML_SHORT_LINK_IMG_1 2.215 0.139 0.480 0.001 score RDNS_NONE 2.399 1.274 1.228 0.793 score HTML_IMAGE_ONLY_12 1.381 1.629 1.400 2.059 score HTML_MESSAGE 0.001 score BAYES_99 0 0 3.8 3.5 score BAYES_999 0 0 0.2 0.2 this would give us lower scores: 5.996 3.043 7.109 6.554
3.7 for a BAYES_999 is a nice default when you start with your setup until it is trained well enough, but later it's a joke
signature.asc
Description: OpenPGP digital signature
