A Dithyramb for Spam

By Paul Ford

An imperfect alternative to fighting spam which no one will implement, but which would be more satisfying than existing proposals.

I receive a lot of spam, like every other Web donkey with a public email address. I've used a bunch of tools to manage the problem, and now I do Bayesian(ish) classification in Mozilla, which sorts and flags a goodly percentage of offers from Nigeria and toner cartridge deals. But last week, after reading another email extolling the magical tongue action of Harvey the Pussy Eating Wonder Dog, I came up with this strategy:

1. We can classify most email as spam or not-spam.
2. The regular action taken by spam filters is to delete the spam, or throw it in a folder for review and deletion.
3. That's fine, but how about if we generate an email, fake an address, and automatically generate a response like “I'm interested! Send me more!” or “Please call me at the following phone number,” or any of several thousand variations on that theme. Then send that message back to the spammer.
4. Lots of spammers don't have working email. But more and more they do, because they're trying to achieve some sort of respectability. Even the Nigerian scam spammers will receive emails now. So a good percentage of them would receive our replies.
5. If from 10,000 - 100,000 people sent automatic, fake replies, those replies would mix in with the far fewer legitimate “I'm interested” replies. The spammers would have to read through all replies and try to extract the ones that were legitimate. The amount of work required to do this would make it impossible to get a good spam con going. Even if they auto-replied to everyone who wrote them, the same thing would happen: more spam in return, more filtering required.
6. The regular action taken by spam filters is to delete the spam, or throw it in a folder for review and deletion.
7. That's fine, but how about if we generate an email, fake an address, and automatically generate a response like “I'm interested! Send me more!” or “Please call me at the following phone number,” or any of several thousand variations on that theme. Then send that message back to the spammer.
8. Lots of spammers don't have working email. But more and more they do, because they're trying to achieve some sort of respectability. Even the Nigerian scam spammers will receive emails now. So a good percentage of them would receive our replies.
9. If from 10,000 - 100,000 people sent automatic, fake replies, those replies would mix in with the far fewer legitimate “I'm interested” replies. The spammers would have to read through all replies and try to extract the ones that were legitimate. The amount of work required to do this would make it impossible to get a good spam con going. Even if they auto-replied to everyone who wrote them, the same thing would happen: more spam in return, more filtering required.

I'm sure there are terrible failures in my thinking, but I like the idea of hammering back at the people who want me to see horses gone wild or buy remote-control mini cars. I want them to work for their money. And maybe they'll get the message that people just friggin don't need more bullshit in their lives.

See also: A Plan for Spam by Paul Graham, a lamb - nay, ram - who jams with spam dithyrambs, can drink a dram, and gives exams from Panam to Siam, for which you must cram. And, for a follow-up, Better Bayesian Filtering.