Pete,

From my own perspective, here are some comments.

1) "unwanted commercial advertising" isn't necessarily spam. I often find rules in the General group from others nominating sources that are generally perceived as being of low value, including by myself, but aren't technically spam by the measures that we use. We consider any first-party relationship with the recipient to be ok so long as they provide an easy method of unsubscribing and they honor unsubscriptions. It would be a good idea to establish a general rule for the service and in handling things that are nominated by customers. You can't be all things to all people, and while advertising is mostly unwanted it is sometimes missed.

2) I don't weight the Gray category on my system. There are too many newsletters and first-party advertising messages that come through for me to give it any weight when such things are based on the provider and not the sender it represents. We do however use hits on this category to defeat some filters on our system (Declude based) because it can be a good identifier of a bulk-mailer and all that comes with it.

3) Moving clean Gray rules over to General will mostly be Ok, but it will definitely cause some false positives with our own more conservative definition of spam. I would actually prefer that you kill the rules with false positives and re-purpose this category as something like "Dirty Bulk" to identify rules associated with specifically dirty lists on bulk mailers. Some are 95% junk, some are maybe 10% junk, and IMO, some like CheetahMail are less than that.

4) I think that 50% legitimate for Gray hits is a big underestimate. Possibly part of that figure is explained by more general rules that target the bulk-mailer and not the source it represents. If care was taken to tag information unique to the sender of the spam newsletter or spam advertising sent through bulk-mailers that are mixed, that will help to segregate the good from the bad. Tagging the From line or the Reply-To line is often a very accurate way to identify these sources, especially when they primarily use tracking links in the body. Some also make use of unique campaign ID's that can be tagged with certain providers. I am not a fan of tagging the provider's own domains because that doesn't allow for differentiation between good and bad, and you must whitelist after each false positive. I would also propose that customer submissions still primarily go to the General category and spam trap hits go into a Dirty Bulk category. I trust the General category less than every other category except Experimental primarily due to the fact that it contains submissions from users with a more liberal definition of spam than I have (I found Amazon.com in there once for instance). I would hope that people wouldn't report this stuff because it can affect everyone on the system.

5) Just an FYI, deciphering whether or not bulk-mail is legitimate, partially legitimate or totally spam is the hardest part of my job, so whatever you do, please don't make it any harder :)

Thanks,

Matt



Pete McNeil wrote:

Hello Sniffer Folks,

 We are reviewing a number of statistics with an eye toward reducing
 false positives. We have already changed a number of our rule coding
 policies where our highest false positive rates are found.

 One of the proposed changes is controversial and I would very much
 like your input about this.

 The Gray hosting rule group currently has a Block-First,
 White-Rule-Later policy. Rules coded into this group are for the
 likes of Constant Contact.

 Some time ago when this policy was drafted the overwhelming
 consensus was that most content arriving from these services was
 unwanted advertisement spam - therefore it was reasonable to
 white-rule legitimate publications as they were identified,
 especially since a single white rule would be shared by all
 subscribers (thus reducing the work and FP load).

 A recent analysis has shown that the situation has changed somewhat
 significantly. In general the following seem true -

 * The gray hosting group typically tags just less than 2% of messages.

 * Of this 2%, approximately half of the hits would be false positives.

 * If this is true then any benefit generated by the group is negated
 by the risk.

 * Also, if a given system does find benefit from the group then that
 benefit would likely be very small.

 If these points stand up to your comments then the proposal is as
 follows:

 - Existing gray-hosting rules with any reported false positives will
 be removed from the system.

 - The remaining gray-hosting rules will be moved to the "ungrouped"
 group (result 63).

 - No special treatment will exist for future rules that might have
 been placed in the gray-hosting group and no special status will be
 maintained for previous members of the gray-hosting group.

 - Result code 60 will be reassigned at a later time.

 ____

 Please let us know what you think about this change. We want to be
 sure that we don't cause any trouble. We would like to implement
 this policy change as soon as possible depending upon your comments.

Thanks!
_M

Pete McNeil (Madscientist)
President, MicroNeil Research Corporation
Chief SortMonster (www.sortmonster.com)



This E-Mail came from the Message Sniffer mailing list. For information and 
(un)subscription instructions go to 
http://www.sortmonster.com/MessageSniffer/Help/Help.html





-- ===================================================== MailPure custom filters for Declude JunkMail Pro. http://www.mailpure.com/software/ =====================================================



This E-Mail came from the Message Sniffer mailing list. For information and (un)subscription instructions go to http://www.sortmonster.com/MessageSniffer/Help/Help.html

Reply via email to