If it is possible for anyone to sit down at his computer and send me email,

and if those computers can be made to run arbitrary software,

then spammers will be able to simulate users sending mail perfectly in order to send me spam, and no technique will be able to stop sufficiently clever spammers.

Many techniques are proposed to stop spam, and they often work in the short term, but all of them are circumventable by clever spammers.

Micropayments. Spammers will use viruses to send spam on innocent user's accounts. They can just as easily use the user's payment credentials when sending spam from his compromised machine. This isn't just a theoretical possibility: already, a substantial fraction of spam is sent from 'zombie' machines -- PCs with a virus that send spam.

Bayesian filters. Spam will consist of innocent text, indistinguishable from something a friend might send me, and a web link. Although less compelling than big colorful spam ads, these will still generate enough revenue for spammers to thrive.

Spammers keep getting better at generating innocent text. At first they just used some random dictionary words, but now they use snippits of real text. I think the next big development will be using text personalized to the recipient. How can they do this? The spammers get many of their email addresses from web pages. On your web page, as well as your email address, you probably have some text which would score as very non-spammy in your Bayesian filter because it has words that are relevant to you. I think spammers will harvest text together with email addresses and seed their emails with it.

Spam Databases Each text will be unique enough to be statistically indistinguishable from regular email. For spams with web links, they can all be made undetectable too, using techniques such as signing up for lots of free web hosting sites. There are an infinity of ways of disguising a web address in an email which motivated spam recipients will be able to decode.

Virus-immune computers. I'm not holding my breath. At the very least, if it's possible to receive software off the network and execute it, lots of users will be fooled into doing this manually.

Physical devices. The only possible way to stop spam would be for it to only be possible to send email from special-purpose "appliances" which run only ROM software and have physically unbreakable security, like some smart cards have. It wouldn't be much fun to use, and it wouldn't integrate well with the rest of your software. Even most physical security devices, like satellite TV decoder chips, seem to get hacked pretty often.

You can't just add a buttonless dongle to a machine, since hacked software will access it. If you add a button to the dongle that the user must physically press to generate a ticket to send each email, then the spammer's virus will be able to invisibly replace the original email text and recipients with a spam message. The only way to make it secure is for the secure device to display the entire message on its screen so the user can check that he's signing the email he meant to send. But can you make a dongle which is capable of rendering an entire email and is still secure?

It would need to render attachments too, since the spammer could replace all outgoing attachments with spam. It would pretty much have to be a general-purpose computer, which means that it will be hackable.

The Big Picture

Presumably there is an equilibrium point in the volume of spam, where the revenue from spam equals the cost of sending it. I fear the point is reached when most people don't bother reading email any more.

Just as no bank vault can withstand a robber with enough time and tools, no mechanism can stop spam. On the other hand, police do seem to be able to cut down bank heists by arresting people. So I'm hopeful that by sending enough spammers to jail, the problem will be reduced somewhat. One virtue of current anti-spam techniques is that they force spammers to do things that are clearly illegal: forging from addresses, hijacking computers, etc.

So far, law enforcement seems to be doing very little to stop spammers. The lesson from the War on Drugs is that if there is a determined producer and a willing consumer, they will find each other no matter how many billions are spent on enforcement. You have to remember that there are a lot of people out there who want to see spams and want buy the products they advertise. These people will respond to spam, no matter how cryptically encoded.

Don Knuth famously declared that 15 years of having email was enough, and gave it up. I have my own milestone in mind: one million spams is enough, after which I'll have to give up email. At the current rate of 300/day, I should reach that point in 2013.

[Oct 20, 2004] I posted the above on June 4, 2004. Since then, dozens of people have tried arguing with me about it, and it doesn't seem likely that I'll change my mind. So I'd like to let this stand as a prediction, and let history prove me right or wrong. I'd love to be proven wrong, of course, and wake up each morning to an inbox containing only things I want to read. So if you have good ideas for how spam can be stopped, please get to work on implementing them!