Re: [Jaws-project-developers] More spam ideas
Status: Alpha
Brought to you by:
ionmx
|
From: Pablo F. <pa...@pa...> - 2005-01-13 21:35:02
|
Hello, > I guess a good approach is to record the IP along with the URLs=20 > contained in the comment. If more than *2* comments being reported as=20 > spam do come from the same IP, then it's a safe bet that it's actually=20 > spam. If more than *5* reported comments do have the same URLs it can be=20 > considered spam even if the IP is different. > Of course those numbers should be fine tuned to get the best results. That's ok, but I have a question: The IP will be of the submitter or of the URL? Wherever the ip comes, I don't see much case. Why? * Lets suppose that the IP comes from the submitter, what makes you sure that there's no trojan/virus in the computer of the user that is spamming your site?. Maybe it's not the user.. it's the computer with a windoze infected. * Now, lets suppose that you mean the IP of the url. That's worst cause there's no warranty that the URL will resolve (maybe its fake).=20 > Even when a post is considered spam the submitter should have the chance=20 > to prove the system wrong. To acomplish this, in the case a comment is=20 > considered as spam the submitter is presented with a turing test to=20 > prove he is a human :) That sounds nice ;-) Before I came to the school I read the reply of Jon. He says that the antispam thingie should work in the URL, that's nice, a BlackList for URL should work, the list will get bigger, but that's a solution. Now, sorry to be so pessimist with my comments. But two years ago I worked in a company building a system in Perl that will send TONS of spam mails to a big list (of 1 giga, one address by line). They have some restrictions to send their spam (thats why I accepted). Some of the things that I built was a nice algorithm in Bayesian that will be 'learning' how to send mails, you know, which mails first, which mails afters.. a nice queue, etc.. So, if I did that, I can't imagine that other companies have more powerfull algorithms than the app I developed. Greetings, Pablo --=20 Pablo Fischer Sandoval (pablo [arroba/at] pablo.com.mx) Cel: (044-55) 2689-6351 Fingerprint: 5973 0F10 543F 54AE 1E41 EC81 0840 A10A 74A4 E5C0 http://www.pablo.com.mx http://www.debianmexico.org |