Readers: Theres No One Solution to Spam

eWeek Labs' Special Report on the challenges posed by unsolicited e-mail spurs variety of responses.

eWEEK Labs Aug. 19 Special Report on spam—titled, subtly, "How to Slam Spam"—was borne of our own frustrations with the sometimes-blush-inducing junk increasingly crowding our in-boxes. As we suspected, we are not alone, and following are just a few of the responses we received to the report.

It Takes a Village

I checked out the maps site, and its points are valid enough. But I also checked out (the BBS response to MAPS) to see what a marketer had to say in response.

From my perspective, I feel that (or any other marketer) could avoid the problem of being a source of spam by a carefully worded agreement with its customers—that is, if you are deemed to be spamming by, say, MAPS, your connection will be terminated.

Another thought would be reverse spamming: a coordinated effort to clog the arteries of spammers. As the e-mail from a known spammer starts to flow out, why not have it trigger an automated response to flood the spammer with unsolicited replies?

Despite the above, anybody can set up a mail server and go nuts. All the legislation in the world isnt going to resolve the issue. However, framing a worldwide anti-spam parameter of compliance for ISPs would not be an unreasonable start.

— Sim Brigden

New Tack

Traditional content filtering will never work, for the reasons you note (constantly evolving spammer tactics and the unacceptable chance of false positives). For a novel anti-spam approach that has some real potential, check out A fellow named Paul Graham has come up with a new (at least to me) approach to spam filtering that is based on statistical probability rather than traditional content filtering and that could potentially render elaborate content filtering services unnecessary.

He assembled collections of spam and nonspam messages and ran them through an algorithm that calculates a "Bayesian combination of the spam probabilities of individual words." He claims that his Bayesian filter misses only five spams per 1,000, with zero false positives. As an example, a message containing an innocent occurrence of "sex" (such as "please send me a copy of the sex offender study") is delivered, whereas porn solicitations are zapped.

Its not a product but rather an experimental concept he has developed to exercise his primary work with LISP. But it seems to have great potential. If it becomes a product, Ill sign up.

— Lance Groth, Director of IT Services, Minnesota Office of the Legislative Auditor