The CAN-SPAM (Controlling the Assault of Non-Solicited Pornography and Marketing) Act of 2003 was controversial from the start. I think it’s fair to say that nobody thought it would solve the spam problem, but many (such as this guy) thought it could help.
Has it solved the spam problem? No, of course not. Has it helped? Yes, marginally.
It has helped in two ways: First, there have been a few prosecutions under the law, basically of high-profile spammers who were also being prosecuted under other fraud-type offenses. CAN-SPAM is, in such cases, at least some leverage for prosecutors. But that’s a very small benefit.
The other big thing that CAN-SPAM did was to set rules for businesses to follow in order to do mass-mailings. These were the most controversial part of CAN-SPAM because they were opt-out instead of opt-in. This is why critics said, and continue to say, CAN-SPAM “legalized spam.” But it did also require that those businesses make opt-out provisions explicit in communications and to observe them, and this is an improvement over the past.
Many will say (and yes, I have seen your complaints) that some businesses don’t follow through on these opt-out requests, but the real spam problem was never this sort of business. The spammers sending the overwhelming majority of the spam out there are not even pretending to comply with these laws.
So what can be done? I spoke with Greg Shapiro, CTO at Sendmail, the prototypical mail server company. He advocates for sender authentication and domain reputation checks. I’ve been on this bandwagon for years, although my recent realization that Webmail-based spam undermines it has dimmed my enthusiasm, and Shapiro acknowledges these problems. They need to be addressed through proper network etiquette, which means that ISPs and other mail providers, including Webmail providers, need to rigorously monitor outbound mail through filters and network pattern analysis to look for spamming behavior before it leaves their network.
These aren’t perfect measures by any means, not that I have better suggestions. In fact, I assume that Webmail providers are doing what they can now to enforce “network etiquette” within the bounds of cost and other considerations. The real problem they have is that their CAPTCHA tests for account signup can be hacked and scripted so that attackers can create enough accounts to live within the bounds of network etiquette.
Years ago I even toyed with the idea of replacing SMTP altogether and dismissed the idea as impossible, no matter how great the benefit. Now it seems that the benefit wouldn’t be so great after all. (Good thing they didn’t listen to me …)
A Replacement for CAPTCHA
So Shapiro and I talked about what might replace the CAPTCHA. One possibility is a variation on the old PGP (Pretty Good Privacy) scheme. PGP advocates are anxious to point out that their system provides solid enforcement of identity, so that you can prove that messages are from whom they purport to be from and that they have not been tampered with. But PGP is, almost by design, difficult to work with.
Imagine a new class of e-mail account, a verified account, sort of like EV-SSL, which provides enhanced verification of identity for SSL applicants and a visual feedback, in the form of the green address bar, for Web sites. In this case, before getting one of these accounts, a user would have to get a digital certificate from a trusted certificate authority. There might be many of these, as there are with SSL certificates, or there might be just one: There was some talk years ago about the U.S. post office getting into this business. Anyway, stick with me on this.
To get one of these new accounts, you have to present your certificate to the e-mail provider, who might also act as a go-between to obtain it for you. Every message is signed by some hash of that certificate, which therefore must either reside on your computer or be network-accessible in a secure way.
The result is a PGP-like system where individual senders can be uniquely identified. If enough users adopt this system, then recipients can start to use it for reliable whitelists and blacklists.
I know I’ve ignored a ton of technical stuff, but assume I’ve solved the technical problems. There remain huge problems of privacy, ease of use and so on. What identity information should the certificate authority require from me, an ordinary Internet user? It must be strong enough to ensure identity, yet not onerous enough to violate privacy or be too troublesome to bother with. If I lose my certificate, I’ll need to reprove my identity to the CA in order to get my certificate reissued.
To make sense, does the system require that all users get no more than one certificate? There’s a good argument for this, although it ends some aspects of anonymity, since no matter what address I use I can be identified, if not by name, then as a particular unique individual whose public key may be identical to one used by other e-mail addresses.
I could go on with the problems that such a system presents; as with replacing SMTP, even if you could make it happen technically it won’t happen because people would rather put up with the problem than deal with the solution.
Five years after CAN-SPAM, the spam problem is, in many ways, much worse. CAN-SPAM is not at fault, nor could a “better law” have done any better. As long as there is money to be made by sending e-mails, rules won’t get in the way of the spammers.
Security CenterEditor Larry Seltzer has worked in and written about the computer industry since 1983.
For insights on security coverage around the Web, take a look at eWEEK.com Security Center Editor Larry Seltzer’s blog Cheap Hack.