IETF to Look at Spam

m00nun1t writes "CNET has an article about the Internet Engineering Task Force (IETF) looking at what they can do about spam. According to the article, many of the proposals seems to "require changes in basic e-mail technology", which presumably means SMTP (and about time!). Maybe they are looking beyond just SMTP - anyone have any insights here?"

Comments

Spam can be avoided without protocol changes.

[Krapangor]

You just grab spam by its very nature.
Spam is highly redundant commercial advertisement. And we don't want it. So the basic approach would be to exploit this redundancy to filter from the original message streams. However highly localized approaches like personal mail filters will always fail due to the high variety of spam.
How would your personal mail filter know Japanese or Chinese spam ? Not at all, it would just let the spam mails through.
So we can only solve this problem by a worlwide highly sophisticed effort. And the key component will be again the identification of redundancy.

The idea is to build up redundancy signatures of email messages. The global network of SMTP servers exchange these signatures by a grid orientated P2P network. When the distribution of a special signature is worldwide too high then it's identificated as spam and destroyed everywhere [1]. Note that a distributed annealing algorithm can do this in O(log(log(n)))+theta(n) operations (theta(n) a blotzmann distribution due to the annealing process).

Building up these signatures is a little more complicated, something like MD5 checksum won't be sufficient. After some preparsing we have to project the text into a suitable banach space for further operations. Cleckerson and Allman have proposed an infinite dimensional Lie group for this task, surely any Diffeomorphism group over a Hardy space will do [2]. In this space we can build unique (!) singnatures using a simple frequency domain transformation. However we can't of course exchange an infinite amount of data, so an approximation will have to do it. Reseach any Valdpornik and Rosé has shown that there are decent approximations for this task [3]. So we can use these thing to build up our signatures and the whole stuff works indeed.

However the highly numerical approach requires DSP coprocessors in mail servers. But I think that Intel and AMD will throw in a DSP core to their server processors in a few years anyway, so that's not a big issue.