Blacklists End? Further Rise of the Bayesian?
From an item found on MSNBC comes concern about new spammer tools that can confuse filters using blacklists of spammer addresses...“The feature allows spammers to make their zombie-sent e-mail appear as if it were sent directly from an Internet service provider's systems. Since it's not feasible to filter out an entire Internet provider's e-mail, the new SendSafe program foils the entire blacklisting system”.
Does this negate Blacklisting services? Good thing most filters provide layered email message analysis. Looks like more of the “Is-It-Spam” load will be carried by other filtering tactics like Bayesian analysis, weighted word lists, heuristics, etc.
This makes Bayesian training of a company’s email filter even more important. The tweaking of Bayesian tokens to match a corporation’s unique needs requires an administrator’s diligence. Bayesian training is done frequently when a filter is first installed but tapers off as a better picture of corporate email is programmed into the Bayesian token database.
I’ve often found it amusing to imagine a legitimate company that deals in “small OTC stocks” or even areas of legal “adult entertainment” having all of their email quarantined because a badly trained Bayesian filter tags all of their business email as spam.
It’s the opposite of 99% of everyone else with email but does demonstrate how every business’ email in unique.
One man’s spam is another man’s new OTC order.
AAS
0 Comments:
Post a Comment
<< Home