One of the reasons I started using Thunderbird was its (at the time) excellent Spam filters. After a few clicks, it identified 80 or 90% of the incoming Spam and filed it away.
Unfortunately, the spammers fight back. I was reminded of this last night when I checked my private email for the first time in 36 hours – 187 messages of which exactly 13 were messages that I might actually want to read.
Roughly half of the rest were identified by either Thunderbird or SpamAssassin as Spam. But the rest — some 80 messages — were advertisements for penny stocks disguised as fudge brownie recipes; neither tool recognized them as Spam.
Maybe this is the wrong approach. If 90% of all email is spam, then maybe we should just assume all email is spam, unless proven otherwise. Maybe it is easier to identify real mail than spam.
I am going to try a little experiment. I am going to reset the training data of Thunderbird’s junk mail controls and reverse the usage. I’m going to call real mail spam. So the “Junk” folder becomes the receptacle for potentially interesting email. Seems perverse, but real mail isn’t trying to cloak its identity.
Has anyone tried this approach? Anyone want to try it with me and compare results….?
Sehr interessante Idee! Bin äusserst gespannt auf das Resultat.
doof an der sache ist, dass man das problem anscheinend nicht an der wurzel packen kann.
fast der gesamte mail-verkehr im netz ist mittlerweile spam, aber solange der anteil von mail am gesamtverkehr nur 10 bis 15% beträgt, scheinen die provider kein wirksames mittel finden zu wollen.
I don’t think Bayes cares what you label its categories.
hannes,
This is probably true, but there is a difference. Which messages do I gives to the Bayes filter, the good ones or the bad ones? At the moment I’m giving all the bad ones that it doesn’t catch automatically. This doesn’t seem to be helping.
The characteristics of the good messages don’t change much over time. The bad ones are constantly changing to fool the filters.
Markus,
Ich glaube, dass ist nicht so einfach, wie es tönt.
Vielleicht könnte man port 25 (SMTP) gründsätzlich blockieren, alle E-Mails wäre dann über Servern des jeweiligen Providern kanalisiert.
Bei meiner Schwiergermutter macht T-Online genau das. Die Rucksender-Addresse wird auf die des Abonennten-Halters umgeschrieben. Das war zwar sehr mühsam, aber vielleicht ist es ein Anti-Spam-Massnahme – eine Lösung mit Vor- und Nachteile.