The Long Dark Tech-Time of the Soul

This is a technology focused blog that describes my trials and tribulations with techonlogy which, no matter what brave new world is promised to be just around the corner, nearly always fails to live up to expectations.

Tuesday, February 01, 2005

Unpatent this!

This may be a dumb idea I think it needs to be published, if it hasn't been suggested already. Since if I publish it then its unlikely it'll successfully be patented I figured I would. I call this process "the unpatent", as they say, information just wants to be free. So here goes...

The latest fad in generating spam seems to be sending email from domains that either haven't yet, or have only just been registered, and then terminating their registration. Also a lot of spam will point you to a website that is from a similarly fresh, or as some would call them, unborn domains.

Therefore why not create an email and browser extension that warns the user then they are about to visit a website or read an email from an IP address that is either a) not attached to any domain, or b) attached to a very recently registered domain. Also add the option in case a) or b) to validate that its IP address against some database (similar to the RBL lists used to combat spammers) and not visit it. Such a database could be self-populated by sending in the IP addresses and domain names harvested from emails identified as spam by the current collection of Bayesian filters that are in use. Some kind of scoring system would separate wheat from chaff in the database and let users set thier own threshold of tolerance for recently spawn domains.

If such systems were widely deployed it would make the identification of spam and rogue websites much faster since Bayesian filters are widely deployed and could much more rapidly put IP addresses and domain names into the database than occurs with the RBL databases. Since spam is now a necessarily inefficient process spammers have to send millions of messages to get any significant response which takes time even with fast links. During that time you can guarantee at least someone will have recieved and filtered the spam with their current Bayesian filters. Therefore if this occurs within the first few minutes of the spammers deluge, the rest and be effectively rejected at the SMTP level, and all the website addresses they refer to also filtered out for the population at large.

There are two claims:

1) Spam can be reduced by filtering email and websites based on their use of raw IPs or domain names that resolve to recently or not registered domains.

2) Spam can be stopped early in its sending by using feedback from standard Bayesian filters at the client side and maintinaing a CDDB like database of such IP addresses and domains contained in the spam email.

I look forward to someone implementing such an extension in Firefox and Thunderbird.

0 Comments:

Post a Comment

<< Home