Excellent idea, Pete. Don't worry about catching the first image, just at matching future images against he first one reported.
Darin. ----- Original Message ----- From: "Pete McNeil" <[EMAIL PROTECTED]> To: "Darin Cox" <[email protected]> Sent: Thursday, March 17, 2005 5:54 PM Subject: Re[2]: [IMail Forum] Block jpg gif SPAM On Thursday, March 17, 2005, 2:19:15 PM, Darin wrote: DC> Best current means is by blocking individual filenames. I think there have DC> been one or two furtive attempts at image recognition, but the processing DC> power needed for this to handle the onslaught is currently prohibitive. OCR DC> would probably fall into the same category. Hopefully that will change in DC> the not to distant future. I'm just skimming so I'll be brief. Sorry for being late. In Message Sniffer we are generally successful by attacking the structure of image spam as the message usually carries a number of "fixed points" in a flexible pattern carrying the image. This requires some complex heuristics so I'm not sure how (or if) to accomplish this directly in IMail, but it is possible. As for image recognition - we are working on image classifying technology that will also work... For example, in most cases a hash of the image file will suffice since spam is often distributed en mass. Where image obfuscation is employed we have other fast image classification mechanisms which are capable of producing "loose matches" between obfuscated images. These are not yet deployed, but they will be in the next 18 months if all goes as planned. In the mean time, capturing patterns in the carrier message is usually quite effective. HTH, _M To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/ Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/ To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/ Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/
