First response to the list. I've been lurking since I started using Razor.
On Wed, Feb 18, 2004 at 11:28:06AM -0900, John Andersen wrote: > On Wednesday 18 February 2004 08:29, Mike Burger wrote: > > > Then, once you've cleared out the ham, you could use razor-report to > > report the entire batch of messages in one shot. > > > If enough people send in the same spam, it winds up in the database, and > > the next time it comes through your system, razor will spot it. > > But its ALREADY been flagged as spam by spamassassin. So what's > the point? > To add to the community effort? SpamAssassin (what I use as well) relies on multiple attributes to determine the Spam status of a message (which I'm sure you know already). Razor is just one of them. The fact that you can weight different attributes to your liking makes it more versatile. Razor's (or another hashing db) strength lies in adding another source of "yeah this is Spam" to your existing filters. > At the risk of earning the wrath of this list, I'd like to point out that the > Emporer has no clothes! ;-) > The Emperor has clothes. They just appear to be made from polarized cloth that may, or may not, be transparent based on your viewing angle. > If Razor relies on people submitting known spam, and the overwhelmingly > vast majority of those submitting spam do so from automatically trapped > spam, then razor by definition can never come close to the effectiveness > of those other automated detection agents, as it will always be behind > the power curve. > Good point. But that's the reason I make a point of manually submitting the missed UCE to Razor. I'm hoping it will get added, and thus caught by the Razor check. Case in point - the series of "carders.org" (and similar) spams I've been getting. They keep making it through the bayes filters - even with added training. But even behind the curve, Razor would be a useful tool. > Now for truly MISSED spam, this might make some sense, but its > far better to train your own bayes filters (which will have immediate effect) > than submit it to razor in the hopes that someone else will also and slowly > its razor score will rise to the point of being labeled as spam. > I disagree here. It'll have a greater -personal- effect, but it won't help the community. See my example above. If the spams I'm getting aren't widly distributed (I get them, you don't, let's say) then adding them to Razor will be beneficial to you and the rest of the community. It's easy enough to do both, eh? > But for things already trapped by spamassassin (or any other tool) > feeding razor seems rather altruistic, it will never benefit the submitter > but it might help OTHER people who ONLY uses razor. More effective > help for those other people would be to get them a more effective > spam filter and save the bandwidth of submitting spam to razor. > There's nothing wrong with some painless altruism. It doesn't add any real cost to me to help you here, so why not help? Even better, with a particular UCE having a very low spammieness, and possibly resembling some Ham, it'll take several instances to train your bayes filter. If enough people submit to Razor after the first round to flag it as Spam, you'll gain the benefit from Razor before another instance m,akes it through your filters. > Ok, I'm in my bunker, - - Fire away. ;-) > Nah. Just a polite disagreement. > -- > _____________________________________ > John Andersen > Bagheera Resident BOFH Stormcenter Networks > > ------------------------------------------------------- > SF.Net is sponsored by: Speed Start Your Linux Apps Now. > Build and deploy apps & Web services for Linux with > a free DVD software kit from IBM. Click Now! > http://ads.osdn.com/?ad_id56&alloc_id438&opĚk > _______________________________________________ > Razor-users mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/razor-users
pgp00000.pgp
Description: PGP signature
