Prospective StudentsCurrent StudentsBusinessesCommunityDistanceA to Z Index
TSS Today
Home Account Search
SPAM filter and false positives update

The white list configuration for Blackboard appears to be working correctly.  Let me know if you see messages from Blackboard that are still being incorrectly tagged as spam.

 

The process that moves tagged messages (messages that have [SPAM] in the subject line) is independent of the spam filter process.  When you forward a tagged message to a coworker it will be moved to their Junk E-mail folder, even though it does not pass through the spam filter software.  Remove the [SPAM] tag before forwarding messages.

 

Messages sent to other users on our email system are not processed by the spam filtering software.  In every case that this has been reported, either the message was sent to the recipients old @ctc.edu address, or the message was sent with a [SPAM] tag somewhere in the subject line.

Please check your contact list entries to make sure you are using local addresses for local users.  Note that @ctc.edu addresses are targeted by spammers independently of the @shoreline.edu address.  If you don't need it anymore you should consider having it removed.  This could cut down on the amount of junk that your receive.

 

I wish to thank the folks that called or email me with their specific issues.  I can't fix stuff if I don't know that it is broke.  I would also like to thank the folks that helped me to test the fixes.  I can't always duplicate your environment to determine if the problem is really fixed or not.

 

garyk@shoreilne.edu

SPAM word lists filters

We have a comprehensive word list that we have developed over time.  In all likelihood, any word that you would want to add to the filter is either already there, or has been purposefully excluded due to unexpected results.  Words that are contained within other words or words frequently used by various disciplines result in large volumes of false positives.  Word lists development is a trial an error process that is highly prone to errors and ultimately is the least effective of our filters, accounting for only one half percent of the total hits for spam email.  They are easily bypassed by inserting hidden tags between the characters in HTML messages.  In many cases the words are embedded

in images which cannot be scanned by our filter.  Our key word filter does not scan email headers so it cannot find offensive words in the to: and from: fields.  You can build custom word filters utilizing Outlook rules (which can scan email headers) but this is not for the faint hearted.  Outlook rules can produce unexpected results that may not justify the time invested.  Instructions for creating a rule to handle your [SPAM] tagged email can be found at the link below.  This rule can easily be modified to operate on a word list of your choosing.

 

http://intranet.shore.ctc.edu/intranettss/CREATING%20A%20RULE%20-%20SPAM2.htm

 

(note: URL may be wordwrapped)

SPAM filter false positive issues

The recent total failure of the spam filter system was due to corruption of the Bayesian filter database.  This particular filter utilizes a complex algorithm to learn the e-mail habits of the system.  This filtering technology becomes more accurate over time, as the database matures.  Previous to the failure, this filter was the most reliable of the half dozen filters that are employed.  Unfortunately the Bayesian filter database was corrupt beyond repair and now is in the process of rebuilding.  A process that can take months to achieve a high level of accuracy.

 

It would appear, judging from some of the results that I am seeing, that some of the "white list" database was also lost. (white list is a list of source addresses that should not be filtered).  It is difficult to tell since the failure of the Bayesian filter database could produce similar results, and the person who actually performed the reinstallation is now out on leave.  I have been communicating with Ann to address the Blackboard issue specifically. 

 

Normally messages sent from one local Exchange user to another do not pass through the filter.  I have investigated several cases of locally processed false positives, and in each case the messages were send from outside the system.  Typically the sender was using old @ctc.edu aliases for the local user or they are using a POP type email client.