While weeding out the daily spam, I chanced a look at one that contained a perfectly well readable German advertising message and the obligatory link to click on, but beneath that, the following:
Now, being the sort of person that constantly hopes for hidden meanings in randomly capitalized words, irregular bolding/italicizing, acrostics and all manner of textual irregularities (though so far in vain), this piqued my curiosity, but all I’ve managed so far is to conclude that it probably isn’t a simple Caesar Cipher.
So, I turn to the puzzle-solvers on here, with the caveat that this might not be a puzzle at all, but merely random nonsense (in the interest of full disclosure, though, it’s the presence of the smilies that somehow seems at odds with random nonsense to me).
How does that work? All the keywords I would guess a spam filter could react to are in the actual body/headline of the mail, completely unobscured; the quoted bit was merely tacked on.
Some of the time, it isn’t just based on ‘do these terms appear’ but ‘do these terms appear A LOT, compared to the length of the text?’ So adding a bit of ballast that won’t trigger helps tip the balance.
Too small sample to say, I suppose, but the letter distribution doesn’t look random. Of course, that could be because the random text generator throws in a lot of vowels to make it look more like real words. I tried playing with it in a substitution code cracker, but if the plaintext is indeed German it’s going to be tricky.
Many spam filters use a weighting algorithm to judge whether words are good or bad, and they keep a list of words along with their good/bad scores. You have to train the filter by correcting it when it’s wrong. It then learns to score words like “penis” as bad, and words like “TPS reports” as good. The spammers probably assume the nonsense words won’t be in the known word list, so that could increase the ratio of good or neutral words versus spammy words. I think most spam filters have probably been programmed to compensate for the nonsense word trick by now, it’s not anything new.
ISPs also use spam filters at the mass level to determine if a large number of the same email is being sent. Random junk is put into the emails to make each one unique.
They compensate by ranking random nonsense as zero and any spammy terms as negative, so an email with a ton of neutral text and a single spammy term gets a negative score and is sent to the spam box. Read “A Plan for Spam” for a good, non-technical introduction to modern spam filtering.
I’ll have you know my mother is a fine, upstanding, Christian woman, you heathen blowhard! And if you say that again, I’ll have to ask you to step outside.
So, I guess that means it’s not a secret ‘Help! I’m trapped in a spam factory!’-message?
Well then. Thanks for the replies, everybody; I guess the thread may as well be closed now.
