I strongly agree with your statements on this. Surbl is a good tool but 'John Anderson' does not need to respond to every post with "Don't use razor... use surbl.org," especially since this is a list for razor.
Since implementing both razor and surbl my efficiency ratings have really increased into a range where spam filtering looks viable. Before then, the only mechanism I had to increase the effectiveness of spam assassin was writing my own static regexps which was a very time consuming limited approach. I see a lot of potential in what razor is trying to do and surbl seems to be a good approach. However, this is a constantly evolving playing field and I think each approach will be needed in order to have any sort of long term viability in spam filtering. Joe -----Original Message----- From: Tom Allison [mailto:[EMAIL PROTECTED] Sent: Wednesday, April 21, 2004 6:38 PM To: John Andersen Cc: [EMAIL PROTECTED] Subject: Re: [Razor-users] Razor Agents and Future. John Andersen wrote: > On Thursday 15 April 2004 16:08, Vipul Ved Prakash wrote: > >>I am aware that Razor Agents are not very effective at this point. There >>are a couple of thing I want to point out. First, Razor2/SpamNet, as a >>technology, is quite possibly the best spam filtration system that exists >>today by the virtue of its realtime feedback and high granularity of >>categorization. > > > But how well can this granularity hold up when razor is being fed by > hundreds of automated (SpamAssassin) agents? It can be no > better than the sum of its parts. > > Conversly, what happens when all thes automated feeders stop? > Nobody has time to inspect spam with "eyes only", which was your > original intent (as posted here on this list in the very beginning). > > One possible improvement might be to work with the project at > http://surbl.org/ which holds promise by cutting off spammers from their > source of revenue, by useing the very URLs they foist as weapons > against them. > > And that technilogy is similar in nature to what Razor tries to do, and > requires a DNS like approach that could just as well be handled by > an adaptation of a razor engine. > > 500 different message bodies take a while to get into razor. > But when they all carying links that point to a specific IP or subnet > the meer appearance of said link is a very good indication of spam. > John, I suspect you are trying to sell surbl.org in spite of everything presented to you as the #1 solution. The Theory of razor is impeccable. If it weren't then why would there by things like pyzor and others to try variations on the theme. As for your arguement about the spam assassin clients being so good that Razor isn't necessary. It's bunk. It's not the sum of the parts. You have to think of this as a probability expression. When you are using razor to provide a common point of reference for spam, you are effectively combining every spamassassin out there. So if everyone has a 97% accuracy rate for detecting spam then you have have to pass the 3% error rate on yours and the 3% error rate on mine. That puts you at 0.09% probability of delivering your spam. Add that up to 1000 spamassassins and you have something. However, it's not that clean. Spamassassin is limited by it's approach. The static portion of their scoring (regex patterns) is very clear but quickly out manuevered by the spammers. The Bayes portion is more adaptable (hard to out manuever) but much less absolute in it's determination. So well crafted spam may have a very good chance of getting around many SpamAssassin installations. Where razor starts to shine is this: First, I use razor, but I do not use anything like SpamAssassin. I use bogofilter which is a much more advanced form of statistical filtering than Bayesian statistics would provide. More importantly, it's different from SpamAssassin which gives me a better chance of detecting the spam that has been specifically crafted for SpamAssassin to miss. If I razor-report it, then all of the SpamAssassin clients are aided. In addition to this, if a small group of people razor-report the spam, then the rest don't have to worry about it. The delivery of spam is limited to only those who couldn't detect the spam locally and chronologically before razor agent had confirmation on it's spam status. This surbl.org may be a great tool, and I'm not one to knock it since I haven't tried it. But you're playing of game of "product A sucks, but product B rocks". If anything, advocate what rocks on it's own merits. I am encouraged by Vipul's statements about turning some attention to razor agent. But I'm also discouraged that it's been ignored so badly. What guarantee do we have if the next big money making project comes along to distract the development progress? Unfortunately while the code might be Open Source, the documentation behind it is so lacking that obscurity seem to be a major entry barrier into any improvements being made independently. ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click _______________________________________________ Razor-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/razor-users ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click _______________________________________________ Razor-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/razor-users
