You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Robert Menschel <Ro...@Menschel.net> on 2004/02/28 20:30:25 UTC

Re[2]: How about "Confidentiality assured"

Hello Loren, others,

Saturday, February 28, 2004, 12:20:37 AM, you wrote:

>> > I just got a spam about diplomas which didn't get caught and
>> > noticed the phrase "Confidentiality assured" in the email.

>> Well,  I can think of several reasons why confidential and
>> confidentiality etc may be used in legit emails.
>>
>> I receive many emails with confidentiality disclaimers attached
>> at the bottom. I'm sure you may have seen a few. Disclaimers
>> that say the contents of this email are confidential blah blah blah.

>> I know a rule based on those phrases would be give me a huge
>> number of false positives.

LW> A rule looking for "confidential" would certainly cause problems.  I think I
LW> have yet to see one of those that includes "Confidentiality assured", or
LW> even "Confidentiality".  A rule for the first one is unlikely to cause a
LW> great deal of problem for anyone other than spammers, and even the second
LW> one could be helpful if given a fairly low score.  I believe someone ran
LW> those both through a corpus test, and they got very low, if any, ham hits.

Don't know if mine are the results you're talking about, but,
a) my "confidential" phrase rule does hit ham, and so I've had to be
careful with its score (1.584 of 9, equivalent to 0.880 of 5). I found NO
ham matching "confidentiality assured", and so can score that twice as
high.

body      RM_bpn_Confidential    /(?:total(?:ly)?|VERY|strictly|high(?:est|ly)?|utmost) CONFIDEN(?:ce|T(?:AI|IA)L)/i
describe  RM_bpn_Confidential    says this is very confidential
score     RM_bpn_Confidential    1.584  # 409s/6h of 97268 corpus (79437s/17831h) 01/24/04
                                        # ham: membership list, survey confidentiality, 
body      RM_bpn_Confidential2   /\bconfidential(?:ity)? assured/i
describe  RM_bpn_Confidential2   says this is very confidential
score     RM_bpn_Confidential2   3.000  # 616s/0h of 106556 corpus (87320s/19236h) 02/27/04

Bob Menschel