Re: Telling BAYES not to learn?
Martin Gregorie skrev den 2013-02-11 16:41: Maybe there's a case for classifying mail as ham/spam by reading the raw mail instead of looking at it with an MUA and being shown the HTML part? why is it needed ?, if mua clients dont trust html, then use text mode mua, problem is gone well it is possible, but create an sed -e /html/text/ is a bit overkill, and this stops relearning into bayes, best option is to ask for abuse support in mail clients to not trust any html collers, or redefine them so its basic custom designed to not being what html was ment them to be displayed as :)
Re: Telling BAYES not to learn?
On Thu, 2013-02-14 at 13:18 +0100, Benny Pedersen wrote: Martin Gregorie skrev den 2013-02-11 16:41: Maybe there's a case for classifying mail as ham/spam by reading the raw mail instead of looking at it with an MUA and being shown the HTML part? why is it needed ?, if mua clients dont trust html, then use text mode mua, problem is gone Exactly. It seems to me that a lot of people, possibly OP included, may not realise that most graphical MUAs default to showing the HTML part so the ham/spam reviewer may not realise they are being shown some white-on-white text and/or that the plain text part is often quite different from the HTML part. I don't know any MUA that will show both plaintext and HTML, which is why I suggested that the ham/spam classification reviewer do it with less rather than an MUA. Martin
Re: Telling BAYES not to learn?
Den 07-02-2013 17:13, Marc Perkel skrev: Because when a message uses invisible text to poison bayes then I don't want to learn that because it will make bayes less effective. that still does not make sense, if you say an hidded word is spam then its spam, no matter if both background and forground colors is white, so is bayes but i can agree if you say that in 1000 ham and 1000 spam mails this problem is then its bayes poison maybe time to make clamav rules for hidded spam content ? :=) sigtool --html-normalize spammsg now sigtool have created 2 new html files, that is more clean to make rules on
Re: Telling BAYES not to learn?
On Mon, 2013-02-11 at 16:00 +0100, Benny Pedersen wrote: Den 07-02-2013 17:13, Marc Perkel skrev: Because when a message uses invisible text to poison bayes then I don't want to learn that because it will make bayes less effective. that still does not make sense, if you say an hidded word is spam then its spam, no matter if both background and forground colors is white, so is bayes but i can agree if you say that in 1000 ham and 1000 spam mails this problem is then its bayes poison Maybe there's a case for classifying mail as ham/spam by reading the raw mail instead of looking at it with an MUA and being shown the HTML part? Martin
Re: Telling BAYES not to learn?
On Tue, 05 Feb 2013 07:20:24 -0800 Marc Perkel wrote: is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? On 2/7/2013 6:58 AM, RW wrote: Why do you think this is a good idea? On 07.02.13 08:13, Marc Perkel wrote: Because when a message uses invisible text to poison bayes then I don't want to learn that because it will make bayes less effective. well, I found bayes poisoning not an issue long ago... -- Matus UHLAR - fantomas, uh...@fantomas.sk ; http://www.fantomas.sk/ Warning: I wish NOT to receive e-mail advertising to this address. Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu. Windows found: (R)emove, (E)rase, (D)elete
Re: Telling BAYES not to learn?
On Tue, 05 Feb 2013 07:20:24 -0800 Marc Perkel wrote: is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? Why do you think this is a good idea?
Re: Telling BAYES not to learn?
On 2/7/2013 6:58 AM, RW wrote: On Tue, 05 Feb 2013 07:20:24 -0800 Marc Perkel wrote: is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? Why do you think this is a good idea? Because when a message uses invisible text to poison bayes then I don't want to learn that because it will make bayes less effective. -- Marc Perkel - Sales/Support supp...@junkemailfilter.com http://www.junkemailfilter.com Junk Email Filter dot com 415-992-3400
Re: Telling BAYES not to learn?
On Thu, 07 Feb 2013 08:13:54 -0800 Marc Perkel wrote: On 2/7/2013 6:58 AM, RW wrote: On Tue, 05 Feb 2013 07:20:24 -0800 Marc Perkel wrote: is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? Why do you think this is a good idea? Because when a message uses invisible text to poison bayes then I don't want to learn that because it will make bayes less effective. But those emails are still going to be scanned by Bayes. If a word is being commonly added by spammers there's no point in pretending that it's still a strong ham indicator, such tokens need to be learned as spam so they get detuned and drop-out of the final calculation.
Re: Telling BAYES not to learn?
On 2/7/2013 11:13 AM, Marc Perkel wrote: On 2/7/2013 6:58 AM, RW wrote: On Tue, 05 Feb 2013 07:20:24 -0800 Marc Perkel wrote: is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? Why do you think this is a good idea? Because when a message uses invisible text to poison bayes then I don't want to learn that because it will make bayes less effective. Invisible text is a problem only for humans, not for machines. So, it sounds as though the problem you're describing relates to reviewing messages, manually (with your eyes), and taking some action as a result. If this is so, why not read the messages in *plaintext*, so you see the invisible text and can therefore act accordingly?
Telling BAYES not to learn?
is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? -- Marc Perkel - Sales/Support supp...@junkemailfilter.com http://www.junkemailfilter.com Junk Email Filter dot com 415-992-3400
Re: Telling BAYES not to learn?
Le 05/02/2013 16:20, Marc Perkel a écrit : is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? Yep - tflags RULENAME noautolearn John. -- -- Over 5000 webcams from ski resorts around the world - www.snoweye.com -- Translate your technical documents and web pages- www.tradoc.fr
Re: Telling BAYES not to learn?
On 2/5/2013 10:23 AM, John Wilcock wrote: Le 05/02/2013 16:20, Marc Perkel a écrit : is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? Yep - tflags RULENAME noautolearn That just tells SA to ignore this rule when determining whether to autolearn. I think Marc was looking for a way to tell Bayes not to learn from the message if the rule hits. -- Bowie
Re: Telling BAYES not to learn?
On 02/05/2013 04:20 PM, Marc Perkel wrote: is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? What causes bayes poisoning? (except bad custom high scored rules)
Re: Telling BAYES not to learn?
Marc Perkel skrev den 2013-02-05 16:20: is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? tflags foo-rule noautolearn
Re: Telling BAYES not to learn?
Axb skrev den 2013-02-05 16:30: On 02/05/2013 04:20 PM, Marc Perkel wrote: is there a way I can put something in a rule that would cause bayes not to learn - such as a rule that detects bayes poisoning? What causes bayes poisoning? (except bad custom high scored rules) +1 worst case is whitelist_from ran...@sender.example.org with default score of -100 :(