The problem is that there's no obvious characteristics that these emails have that make them spam vs non-spam. There's random text with an image attached, how do you know it's not just a friend sending you a picture?
The fact that the spam filter catches some and not others is usually due to non-content related factors. You'll probably find lots of hits to do with RBLs and the like for ones that are caught.
If there was a good set of rules to catch these, I would have created one, but the chance of false positives is high. One interesting thing, I got none of these overnight, so maybe they've stopped sending them right now.
Reporting spam currently doesn't train the bayes filter, however it does put the spam into a corpus that will go to build your bayes filter when we enable it, so it is still useful.
Rob
|