interested about (avoiding) spam... :)

[Mendel]

I'd like to learn about certain things in spam... I get a lots of spam like this: ( I choose this one to show since this is quite short)

hznvyg grassu glumiflorae q2 hazarsusah hnatara

New pha:rmACY - Not a single medical question asked, gu-aranteed or it's
fr*ee.

Being gen-eric, our products are less expensive than the brand name
equivalants. The reason that you can bvy our products for less is that we
dont pay for branding, ma~rketing, advertising, or a costly sales force. We
pass along those cost saviings directly

Parac^odin), Su'perVia-gra or more - 0rder M`eds From Home Now


Hit here (*link removed, I don't want to help them*)

We appreciate your business.





Rob had secured his ruby ring again, and going to the chief's tent he
showed the jewel to the guard and was at once admitted

Is this familiar? Should this be even readable? WTF?

All the spam has this obscure first chapter (usually longer) filled with zibberish, that dont make any sense at all then last chapter again is some sort of unrelated story ( usually longer too) and inbetween there is the actual message with all keywords typoed...

O'm just interested whether this is actually a "product" by some spam mailer company. I'm interested in how much time and money are they investigating in research of the "best" kind of measures to avoid spam detectors etc... Of course most of all I would like a vague idea about who to complain to!

And then I would like to have advice in how to avoid these... they always come from different mail address and seem to come from arbitrary ip address too so it's kinda hard to block!


edit: decided to put the spam in a quote

 

[Humus]

Yeah, that bothers me too. You get spam that survives all filters, but seems to have zero purpose. You can't even figure out what product they are promoting. No link, no nothing. Just some random gibberish.

 

[Zaphod]

They are techniques to try to fool bayesian spam filters sometimes called bayes-poison. The misspellings are there to introduce enough randomness in the 'dirty' words (herbal viagra, cheap software, online university, and so on) and patterns to avoid the flags that would tag the mail as spam in the first place. The longer unrelated gibberish text is there to decrease the ratio of ‘spam words’ in relation to what the filter would see as normal text, thus decreasing the chance that the mail will pass the threshold to be tagged as spam. In addition, the idea is that if you teach your filter that enough of these texts are spam your filter will eventually start removing too much normal mail to remain useful. Thus poisoning the filter.

Edit: Spelling

 

[Briareus]

You could try removing all non alpha characters from your email before feeding it to the bayes net.

 

[cristic]

I have to say that the built-in spam filter of Mozilla Thunderbird does a very good job on sorting out spam and non-spam, even those that use random-gibberish to fool the filters... You might want to try it

 

[Zaphod]

I could add that while the early 'poison' used to be random noise, and then random words, this is quite easily defeated by filters. The most recent variants seem to use abstracts from e-books freely available on the net. In the case of the spam quoted above; the last paragraph is taken from The Master Key by L. Frank Baum - An Electrical Fairy Tale Founded Upon The Mysteries Of Electricity And The Optimism Of Its Devotees. It Was Written For Boys, But Others May Read It. (The same guy that wrote the wizzard of Oz).