Alan J. Flavell wrote:

Going back to some earlier discussion, there was a suggestion that manual intervention and review of the item by the mail admin could be a solution. You know, it's happened more than once that a spam was so cleverly worded that I, as mail admin, was on the point of being taken in by it as genuine mail. In all probability, then, there have been occasions where I really *was* taken in. Equally, there may have been occasions where the mail was genuine but I rated it as abusive, I suppose. I think my review is probably a bit more accurate than the automated checks (spamassassin etc.) implemented in the mailer, but I'm by no means perfect (and I'm *far* more expensive than the automated checks, so I try to minimise the time spent on such tasks), so I think I'd have to say (based on this sample) that mail admin review, while it can certainly be useful, is no "magic bullet".

Yup, there's a narrow, but not empty, grey zone where the admin can spend an awful lot of time trying to decide if something's spam. In my case, the messages in this zone are usually what I call "highly targetted spam", spam that targets the user based on their profession/interests. So my web designer user gets spammed by people running "post-bubble-burst web designer support groups", my landscape architect user gets spammed by a german company trying to sell him the latest in water-feature technology, and my wife gets spammed by baby clubs. It looks a lot like spam to me, but it's so targetted to the receiver's usual mail that both the bayesian filter and I think that there's a chance they may have signed up for it. I usually end up calling the user to ask about the sender, and then explicitly white- or black-list the sender so I don't have to worry about the next message they send.

Thankfully, it doesn't happen very often. But when it does, it's a big time sink.

- Marc

--
## List details at http://www.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://www.exim.org/eximwiki/

Reply via email to