View previous topic :: View next topic |
Author |
Message |
Lawrence
Cadet
Joined: May 20, 2004
Posts: 2
Location: UK
|
Posted: Sun Jun 06, 2004 12:12 pm Post subject: Nonsense spam |
|
|
Hi.
I regularly get spam that contains nothing but nonsense. Here are the contents of two I had today:
Quote: |
dropped nephew parallel summary tie third problem |
Quote: |
contradistinction todd bicker pleistocene scrounge anatomic
sacrificial candle furze bequest carbohydrate
erratic kemp vengeful artery |
Does anyone have any idea why this stuff is being sent?
Thanks.
|
|
Back to top |
|
|
rogerw
Major
Premium Member
Joined: May 11, 2003
Posts: 858
Location: USA
|
Posted: Sun Jun 06, 2004 12:24 pm Post subject: |
|
|
Yes - to try to slip past Bayesian filters (like MW has).
Many big ISPs also use Bayesian filters. The filtering works by assigning good/bad weighting factors to various words that were in the training emails then using the weights to score incoming mail. An email that scores 'bad' is 'probably spam'
Since many spams and good emails have words in common, many words are rather 'neutral'. Words that haven't been introduced into the Bayesian database are weithed to the good side' - but only slightly.
Thus, by adding many words that seem to be randomly taken from the dictionary, the spammers hope to get the score for an email into the probably legitimate range so that it'll not get filtered out. |
|
Back to top |
|
|
Lawrence
Cadet
Joined: May 20, 2004
Posts: 2
Location: UK
|
Posted: Sun Jun 06, 2004 8:18 pm Post subject: |
|
|
Thank you roger. |
|
Back to top |
|
|
rogerw
Major
Premium Member
Joined: May 11, 2003
Posts: 858
Location: USA
|
Posted: Sun Jun 06, 2004 10:07 pm Post subject: |
|
|
De nada.
The sad part is that the tactic might (given a long time) defeat bayesian filtering by 'polluting' the 'junk' word database with random inncuous words. If you add so many junk emails with random words into the database (training), when those same words appear in 'legitimate' emails, those mails might be scored as 'junk'.
On the plus side, since they're random words, they'll not likely pollute the scoring much - and will not likely appear in your email from Aunt Minne - so Auntie's email probably will get through. The weighting factor for a single word is low - and if the same word doen't occur in other of the 'junk' you train with - it'll likely not have too much of an effect.
In the short run (for indivdual emails attempting to elude Bayesian filters by getting their scoring into the 'not junk' range), the tactic seems to work. |
|
Back to top |
|
|
Ikeb
General
Premium Member
Joined: Apr 20, 2003
Posts: 3555
Location: Canada
|
Posted: Mon Jun 07, 2004 12:03 am Post subject: |
|
|
Makes me wonder why SPAMers bother. If someone has gone to the trouble of putting sophisticated filtering methods in place to detect SPAM, what makes them think "if only I can just defeat their filters I'll win them over with my sales message. How can they resist MY promises of manhood enlargements, 1,000,000 bucks, naked nymphomaniacs, the best software/hardware deal ever, etc. etc.?"
_________________
I like SPAM ... on my sandwich! |
|
Back to top |
|
|
AlphaCentauri
Captain
Joined: Nov 20, 2003
Posts: 302
Location: USA
|
Posted: Tue Jun 08, 2004 12:04 pm Post subject: |
|
|
Because your spam might be filtered by someone other than yourself. Your parents may have had your ISP activate a filter. Or the system administrator on your job might be filtering things they don't think you should be spending time looking at on their dime. SOME doofus is answering the stuff, and it's that kind of intermittent reinforcement that really encourages them. |
|
Back to top |
|
|
stan_qaz
General
Premium Member
Joined: Mar 31, 2003
Posts: 4119
Location: USA
|
Posted: Wed Jun 09, 2004 10:19 am Post subject: |
|
|
One buyer in a million spams is a money making return rate when they can send 10 million spams in a few minutes of their time. |
|
Back to top |
|
|
|