random words within html tags
Alvaro Cardenas <[email protected]> Thu, 11 Dec 2008 01:35:48 +0000
| Newsgroups | gmane.mail.spam.spamprobe.general |
|---|---|
| Message-ID | <[email protected]> |
--===============7817797561615425692==
Content-Type: multipart/alternative;
boundary="_93a57d7c-53bb-4a6c-9a64-754976221ccd_"
--_93a57d7c-53bb-4a6c-9a64-754976221ccd_
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Hi all=2C=20
I am a little bit confused about the html parsing that spamprobe does. In =
the main description of how spamprobe works it is said that html tags are i=
gnored.=20
However=2C I am attaching an example of a recent spam email I received.=20
This is included within the <body> of the message:
<STYLE> said Gervasio=2C relate to others and activities plenty of time wh=
en they can Atlanta=2C Georgia.said Gervasio=2C children are plopped in=20
Here's some soothing </STYLE>
and this is the result of spamprobe by using the scoring -T option:
Spam Prob Count Good Spam Word
0.9999990 1 0 2920 padding
0.9999973 1 0 999 0px
0.9999931 1 0 388 men
0.9999926 2 0 360 gervasio
0.0000108 1 4 0 already have
0.0000108 1 4 0 we already
0.9999848 1 0 175 relate to
0.9999847 1 0 174 georgia
0.9999769 1 0 115 margin 0px
0.9999767 1 0 114 0px padding
0.9999749 1 0 106 Hsubject_we have
0.9999749 1 0 106 atlanta
0.9999749 1 0 106 others and
0.9999734 1 0 100 are plopped
0.9999734 1 0 100 plopped
0.9999734 1 0 100 plopped in
0.9999734 1 0 100 to others
0.9999726 1 0 97 plenty of
0.9999704 1 0 90 children are
0.9999704 1 0 90 s some
0.9999704 1 0 90 some soothing
0.9999704 1 0 90 soothing
0.9999694 1 0 87 2px
0.9999675 1 0 82 said gervasio
It is clear that spamprobe is using these "hidden" words as part of the cla=
ssification. =20
How can I make spamprobe avoid using words that do not appear to the final =
user?
I know that in this case=2C spamprobe classifies the message as spam=2C but=
I am not sure it will classify messages correctly the next time.=20
thanks=2C
Alvaro
_________________________________________________________________
Lancez des recherches en toute s=E9curit=E9 depuis n'importe quelle page We=
b. T=E9l=E9chargez GRATUITEMENT Windows Live Toolbar aujourd'hui !
http://toolbar.live.com=
--_93a57d7c-53bb-4a6c-9a64-754976221ccd_
Content-Type: text/html; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
<html>
<head>
<style>
.hmmessage P
{
margin:0px=3B
padding:0px
}
body.hmmessage
{
font-size: 10pt=3B
font-family:Verdana
}
</style>
</head>
<body class=3D'hmmessage'>
<pre>Hi all=2C <br><br>I am a little bit confused about the html parsing th=
at spamprobe does. In the main description of how spamprobe works it is sa=
id that html tags are ignored. <br>However=2C I am attaching an example of =
a recent spam email I received. <br><br><br>This is included within the <=
=3Bbody>=3B of the message:<br><br><=3BSTYLE>=3B said Gervasio=2C rel=
ate to others and activities plenty of time when they can Atlanta=2C Georg=
ia.said Gervasio=2C children are plopped in <br>Here's some soothing <=3B=
/STYLE>=3B<br><br>and this is the result of spamprobe by using the scorin=
g -T option:<br><br> Spam Prob Count Good Spam Word<br> 0.9999=
990 1 0 2920 padding<br> 0.9999973 1 0 9=
99 0px<br> 0.9999931 1 0 388 men<br> 0.9999926 =
2 0 360 gervasio<br> 0.0000108 1 4 0 alr=
eady have<br> 0.0000108 1 4 0 we already<br> 0.999=
9848 1 0 175 relate to<br> 0.9999847 1 0 =
174 georgia<br> 0.9999769 1 0 115 margin 0px<br> =
0.9999767 1 0 114 0px padding<br> 0.9999749 1 =
0 106 Hsubject_we have<br> 0.9999749 1 0 106 at=
lanta<br> 0.9999749 1 0 106 others and<br> 0.9999734=
1 0 100 are plopped<br> 0.9999734 1 0 =
100 plopped<br> 0.9999734 1 0 100 plopped in<br> 0.=
9999734 1 0 100 to others<br> 0.9999726 1 0=
97 plenty of<br> 0.9999704 1 0 90 children are<=
br> 0.9999704 1 0 90 s some<br> 0.9999704 1 =
0 90 some soothing<br> 0.9999704 1 0 90 soo=
thing<br> 0.9999694 1 0 87 2px<br> 0.9999675 =
1 0 82 said gervasio<br><br><br>It is clear that spamprobe is u=
sing these "hidden" words as part of the classification. <br><br>How can I=
make spamprobe avoid using words that do not appear to the final user?<br>=
<br>I know that in this case=2C spamprobe classifies the message as spam=2C=
but I am not sure it will classify messages correctly the next time. <br><=
br>thanks=2C<br>Alvaro<br><br></pre><br /><hr />Discutez gratuitement avec =
vos amis en vid=E9o ! <a href=3D'http://get.live.com/messenger/overview' t=
arget=3D'_new'>T=E9l=E9chargez Messenger=2C c'est gratuit !</a></body>
</html>=
--_93a57d7c-53bb-4a6c-9a64-754976221ccd_--
--===============7817797561615425692==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
------------------------------------------------------------------------------
SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada.
The future of the web can't happen without you. Join us at MIX09 to help
pave the way to the Next Web now. Learn more and register at
http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/
--===============7817797561615425692==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
_______________________________________________
Spamprobe-users mailing list
Spamprobe-users-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
https://lists.sourceforge.net/lists/listinfo/spamprobe-users
--===============7817797561615425692==--