Skip to main content
Welcome guest. | Register | Login | Post

Spam Filter idea

1 reply [Last post]
supermike's picture
Offline
Joined: 2006-02-17

I had a spam filter idea. The way it works is by a sysop posting 15 separate but legitimate questions in 15 separate computer-oriented newsgroups on USENET (or groups.google.com). He leaves a return address of his mail server instead of cloaking it. He asks people to reply in the forum instead of to his email address.

Of course, as expected, the spam starts to flow in. However, it goes into a dead bucket. There, a unique keyword algorithm is applied that removes vowels, numbers, characters, and anything but verbs as far as parts of speech. This leaves verbs and keyword phrases without vowels. It stores this in an indexed database for up to 100 days.

When an end user receives mail on this server, the same keyword algorithm is applied. It then compares it with the dead spam bucket to see if it has a match by a percentage of 80% likelihood or higher. It need not be an exact match. If it finds this, then it deletes the message and sends an alert to a central mail filter server farm used by the community that supports the product. This helps the mail filter server farm issue known spam alerts to all the other mail servers that utilize this technology and to block that spam.

To help the community protect its prized mail filter server farm, proxies are utilized that hide the exact location of these, and these are dispersed on the Internet all over the globe. Meanwhile, a GPG public and private key scheme is used for the communication, along with a CRC check, to prevent packet mangling and ensure that the sender and the recipient are who they say they are.

Meanwhile, a more traditional Baysian spam filter can be utilized. By combining two separate spam filter technologies, it can help make up for the other one.

Next, and this is something I haven't figured out yet, one needs to block the crackers who like to crack such a system and create the kind of spam that goes around this filter. More than likely the net traffic can be analyzed to see if something is being tested, over and over, to see if it's defeating the mail filter. One would need a fuzzy logic script to analyze that kind of pattern and knock that IP address off.

libervisco's picture
Offline
Joined: 2006-05-04
This sounds like a pretty

This sounds like a pretty complex, but potentially more efficient system.. These ideas are precious and you seem to have lots of them..

It would be good to see some opinions on this from experts. Eye

Would it work?

Comment viewing options