Print Page | Close Window

Opinion on Bayes effectiveness

Printed From: LogSat Software
Category: Spam Filter ISP
Forum Name: Spam Filter ISP Support
Forum Description: General support for Spam Filter ISP
URL: https://www.logsat.com/spamfilter/forums/forum_posts.asp?TID=5250
Printed Date: 01 May 2025 at 10:58pm


Topic: Opinion on Bayes effectiveness
Posted By: WebGuyz
Subject: Opinion on Bayes effectiveness
Date Posted: 30 June 2005 at 12:00pm

I feel with the advent of most spammers trying to poison bayes by sending tons of legal words in addition to the trash they peddle, the effectiveness of Bayes has diminished. I am considering no longer using it since the amount of processing power required for each and every email is quite substantial.

I see people on the list writing all kinds of neat regex scripts to try and stem the tide and the realization hits me. What these scripts are trying to do is emulate what Spamassassin 3.0 already does as a matter of course.

I use a spamassassin product behind my Spamfilter box and am amazed at the amount of stuff that gets through. I use this info to create filters for Spamfilter, but what a lot of time to spend.

The beauty of Spamassassin is that the scripts are updated and there are a ton of customizations that people have written to fight a new spam as it comes to light. Also you can manually weight the different types of spam detected to the type that your receiving

Anyone else have an opionion about this?

 

 



-------------
http://www.webguyz.net



Replies:
Posted By: Desperado
Date Posted: 30 June 2005 at 3:35pm
My opinion only:

I will add more later but ... the problem I see with SA is that there are a lot of very "Subjective" regular Expressions ... great for a corporate environment but horrible for an ISP environment like ours where our customers are corporate, not residential

When we were evaluating anti-spam systems, we were already using SA and were getting nearly 35% false positives based on customer complaints.  Our customers would rather get some spam but never tolerate the false positives.  We practice a "No Censorship" policy where SA tends to censor content with very aggressive expressions.  Yes, you can tone it down but I prefer the way SpamFilter allows me to really write my own blocking levels.
 
Regards,


-------------
The Desperado
Dan Seligmann.
Work: http://www.mags.net
Personal: http://www.desperado.com



Posted By: WebGuyz
Date Posted: 30 June 2005 at 3:53pm
Originally posted by Desperado Desperado wrote:

My opinion only:

I will add more later but ... the problem I see with SA is that there are a lot of very "Subjective" regular Expressions ... great for a corporate environment but horrible for an ISP environment like ours where our customers are corporate, not residential

When we were evaluating anti-spam systems, we were already using SA and were getting nearly 35% false positives based on customer complaints.  Our customers would rather get some spam but never tolerate the false positives.  We practice a "No Censorship" policy where SA tends to censor content with very aggressive expressions.  Yes, you can tone it down but I prefer the way SpamFilter allows me to really write my own blocking levels.
 
Regards,

Isn't that what the quarantine db is for. Even if Spamassassin scored it high for whatever reason it would be in the DB and users just check it and get it delivered, the sender gets whitelisted, and its never stopped again.

I have been evaluating competitors products like XWall and they have some good ideas, but I like the basic engine in SF better. Just wish it had few more things like greylisting and spamassassin. I am already using one idea I got from XWall and implementing autowhitelisting of outgoing emails, but worrying about how big that AutoWhiteListForceDelivery.txt file can grow to before it become unbearable.



-------------
http://www.webguyz.net



Print Page | Close Window