DISQUS

DISQUS Hello! Elliott Back's Blog is using DISQUS, a powerful comment system, to manage its comments. Learn more.

Community Page

Elliott Back's Blog

Jump to original thread »
Author

Here’s one way to beat baysian and other content filtering

Started by elliottback · 8 months ago

No excerpt available. Jump to website »

4 comments

  • marcovhv

    3 years ago

    I got many, MANY prescription-meds spam last year using this technique. They just add a load of garbage text and it beats any spamfilter.
    reply
  • Elliott Back

    3 years ago

    What's sad is that this is not even the best way to beat bayesian filters... just a rather clever dumb one.
    reply
  • Steven

    3 years ago

    This is not really true. Get a better Bayesian filter. Words that are not in your corpus as either good or bad will not be used by the Bayesian filter to judge the message one way or another.
    reply
  • Elliot Lee

    3 years ago

    In your screenshot, "Thunderbird thinks this message is junk." Looks like it worked after all!

    And if not, I don't see why bayesian filters don't consider image tags as well. Sure, you might not be able to decipher the text on-the-run; but you know there's an image there, and the size of the image. If it's just a little clipart or icon, it could pass. You could then program your filter to look down on messages that start off with a large image, for example.
    reply

Add New Comment

Returning? Login