Total spam dropped again, to 1299, with 7 misses (99.46%). False positives lept to 36, which can partially be attributed to my laziness in reclassification after upgrading to POPFile 0.19.1. In all, 41.8% of my e-mail for August was spam.
I’m satisified with POPFile 0.19.1’s classification performance, however, it’s overall impact on my system came as […]
Tag Archive for 'bayesian'
August Spam Stats
25Sep03July Spam Stats
03Aug03Spam for the month of July plummeted to 1317. POPFile missed 13 spam messages (99.01%), and there were 17 false positives that included one personal e-mail and 13 from assorted Yahoo Groups mailing lists. In total, 39.7% of my e-mail was spam.
I believe that this is the first time that my spam intake has fallen […]
Upgrading POPFile
31Jul03Today I upgraded POPFile from v0.17.5 to 0.19.1. Sticking with the same old version for eight months apparantly had me missing out on quite a few improvements. Since I started using fetchmail, I’d been bumping into an issue where mail retrieval would never time-out if my Internet connection was down. Plenty of blame to go […]
POPFile 0.18.3 and 0.19.1
16Jul03Yesterday brought us two new POPFile releases (announcement). The changes primarily deal with reliability for email downloads, but the announcement also mentions parser tweaks and the resulting temporary decrease in accuracy. This is why I’m still using 0.17.5 — I reclassify messages very infrequently, so I don’t want to upgrade the parser more often than […]
June Spam Stats
16Jul03My spam stats are always relevant, even after a prolonged period of silence. For the month of June, spam surged to 51.4% of my e-mail. This eerily coincides with some spam statistics published last month.
There were 2112 spam messages, of which 14 were not caught by POPFile (99.34%). There were 17 false positives — one […]
May Spam Stats
02Jun03For the month of May I received 1684 spam messages, and 24 of them made it to my Inbox (98.57%). There were 5 false-positives, one of which was a genuine personal message from a former colleague. In total, 43.6% of my e-mail was spam.
April Spam Stats
19May03On the subject of spam, I’m still in love with POPFile. Last month I received 1182 spam messages, only 23 of which made it to my Inbox (98.05%). There were nine false positives — one spam posting to the POPFile mailing list, three spammy confirmation messages, one tech support e-mail from Belkin, and four bounced […]
Bayes will be defeated?
10Mar03Jeremy Bowers has outlined an attack against Bayesian spam filtering (via k5).
Jeremy’s premise is that, for a given language, everyone’s non-spam corpora will be very similar. Spammers can exploit this liguinstic similarity by building their own corpora, identifying spam words and replacing them with words that have a higher non-spam probability.
Jeremy demonstrates this using a […]
