On Monday, October 4, 2004, 4:47:57 PM, Daniel Quinlan wrote: > Lists of mostly good URLs:
> 1. http://dmoz.org/ - Open Directory Project > huge human-edited list of URLs that are mostly good Good suggestion; the human-edited part makes it useful. > 2. http://www.google.com/ - Google > highly ranked URLs are mostly good, there is an API for > programmatic access Big spammers also get high rankings. http://www.google.com/search?hl=en&lr=&ie=UTF-8&c2coff=1&q=optinrealbig&btnG=Search There are also spammer/scammers who specifically craft search engine submissions to get high scores: http://groups.google.com/groups?selm=2b379589.0406081103.7c29a65e%40posting.google.com > 3. Top N lists of sites: > http://www.pcmag.com/category2/0,1738,7488,00.asp > again, mostly good We have the Alexa 500, Fortune 1000, all publically listed companies on U.S. Stock exchanges, etc. It's not enough, and even some of those are somewhat controversial. Alexa has things like the top 10,000 or 100,000 but they charge money for them, and I assume they may contain some spammers. > 4. SURBL query traffic > mostly good if you subtract the blacklisted ones But any big, as-yet-undetected spam domains can also generate much traffic. > Identifiers and filters: > 1. SBL queries on NS->A records for domain of site, if it's listed, > then it's probably a spammer. Yes, it's in our draft policies for manual lists: http://www.surbl.org/policy.html > 2. Hits 1, 2, or more of the above good lists. > 3. Registration cost: cheaper registrar = more likely to be spam, > more expensive registrar = less likely to be spam. Perhaps not too important. Professional spammers probably make enough money that domain registration fees are in the noise for them, given that they can afford to register hundreds or thousands per day. Arguably more interesting are registrars who seem to host a lot of spam domains, and name servers recognized as controlled by spammers, such as those included in SBL. > 4. Listed in Catherine Hampton's SpamBouncer domain list = almost > certainly spam. We are looking into using SpamBouncer as a data source. > 5. Domain is listed in ROKSO = definitely spam > 6. Domain is listed in Yahoo (similar to DMOZ) = probably not spam > 7. Domain appears in Wikipedia = probably not spam All probably good tests. Will add checking DMOZ, Yahoo, Wikpedia to our draft inclusion policy. ROKSO is part of SBL. > I hope these ideas help. I've tried to focus on databases that can be > downloaded or accessed. > I think your continual nagging of SA corpus submitters for ham hits is > really weak and only serves to damage the SpamAssassin scoring process > and your own evaluation of efficacy. I hope nobody (else) gives you > their SA ham hits, otherwise we'll have no way to know if the above ways > to reduce FPs are actually working for the average user. There's no need to get personal, and you can improve your measurement of spam and ham detection by improving your corpora. For example larger corpora might arguably be more representative of mail in general. > I think we need to care about more than just appearing to be accurate. The goal is not to appear to be accurate. The goal is to remove as many FPs as possible, regardless of the source. Unfortunately ham corpora are somewhat limited, but we still seek every way to reduce FPs including them. Jeff C.
