The Kiwi originator of the Death2Spam (D2S) filter is exploring mainstream business uses for its technology in document classification and “permission” marketing.
Richard Jowsey and his brother Mike, who handles the marketing, are home from their temporary residence in Australia to negotiate with at least one major prospective New Zealand client in the “communications” space, a company they decline to identify publicly.
The Bayesian algorithms on which D2S is based “learn” the characteristics of spam, trashing emails with similar content to the ones the user has identified (see Kiwi spam filter gets few wrong). Equally, Richard Jowsey says, they can be tuned to recognise and classify documents by their theme as part of an internal document management system. He is currently experimenting with such extensions.
Another potential market, he says, is in teaching legitimate “opt-in” marketers how to phrase their communications so they do not get mistaken for spam. An untutored opt-in marketer may use very similar language to that of spammers, meaning the user may miss out on desired messages.
Does this not open the dangerous possibility of spammers learning how to evade the filter? Spammers don’t work like that, Jowsey says, confidently. They put their faith in “just screaming at you again and again”. But he acknowledges there are some fine ethical lines in the use of such a powerful weapon.
The original D2S works on an applications service provider (ASP) basis. It resides on servers, currently in the US, Canada, Australia and New Zealand. The email passes through them and is flagged as spam, good or doubtful. The user then goes to his/her individual account on the D2S site, classifies the doubtful messages and corrects spam identified as good or vice versa.
An SMTP proxy version of the program is now available for organisations or workgroups to place on their own servers.
It would have been “the line of least resistance” to provide an executable file that users could set up on their PCs, Jowsey says, but this presents a number of problems. It means the accommodation of non-mainstream browsers and email clients, and the chore of updating many individual copies to incorporate enhancements and to learn the latest common spam and virus characteristics.
The fact that the user doesn’t have to worry about keeping versions of the application current has always been a classic selling point of ASP, and is particularly applicable to infrequently used applications like spam filters, he says.
Putting the filter on a few major servers also allows users’ identification of spam to be shared, but this is a double-edged sword, Jowsey admits. There is no such thing as a universal opinion of what’s spam and what’s not. D2S has now implemented separate regionalised and individual spam definitions, to avoid one user being unduly influenced by another’s idea of spam.
Surprisingly, concepts of spam in the US, Australia and New Zealand are discernibly different, Jowsey says – a phenomenon which may be discouraging to those seeking international action on spam.