On Sun, 25 Jul 2004 05:31:35 -0700, "Loren Wilton" <[EMAIL PROTECTED]> writes:
> Just reading their summary, I think it is nice research, but not > really useful. The correlation method seems like a good idea. > Except as they point out it makes granite flow look swift. So not > actually useful at this point intime, but it should be kept in mind. Don't be so hasty. I'd be perfectly happy to spend 60 seconds/message, if it could drop the FN rate by 95% and keep the same FP rate as SA with a threshold of 8. A minute of computer time is worth 1% to me compared to 10 seconds of my time, especially if it is done asyncroniously so I don't have to pay an hour delay when my client reads email. Even if a mailserver couldn't handle that CPU load, it still might be a better solution to setup a P2P system on a hundred desktop machines to do the filtering rather than drop the expensive filter. Scott
