Spammer Tricks
Well, maybe not Technology Good, but you have to give credit to the spammers, they can be creative. One of the most recent tools for fighting spammers is the use of a Bayesian filter. Basically it ranks words based on a mathematical (magical?) formal and uses that ranking to determine if a any given email is a spam message (okay it’s a bit more complex, but you get the idea). Well, today I got a message with the following words in it:
* Stop Premature Ejacu52leclation! * Produce Stronger, Rock Hard Erep7217ctions * 100% Safe To Take, With NO Side Effnmrdhects * Fast Priority Shipping Worldwide * Doctor Approved and Recomu8d2tmended * No Pumps! No Surgery! No Ex9sk1jercises!
Basically what the spammer is doing is trying to counter the Bayesian filter by deliberately misspelling words to avoid classification. This points out one of the biggest problems with spam filters, they have to be reactive instead of proactive. They have to counter every new trick that a spammer comes up with in order to be effective, but they can only do this after the spammers have started using the technique. The same thing happens with anti-virus software. In both cases “evolution” will eventually produce stronger, more effective programs faster and faster. The future looks interesting.
July 5th, 2003 at 3:12 pm
This trick will really, really not work. Misspellings like that are very uncommon in non-spam, while the correctly-spelled words are more likely to appear; so misspellings cannot but increase the spamminess of the mail from the POV of Bayes.
What does cause problems with Bayes is spammers who append chunks of legit mails to the end of their spam; in that case, you need something other than a bayesian scanner to identify it.