Mailman 3 December 2002 - SpamBayes

Outlook plugin - training
by Moore, Paul Dec. 3, 2003

Dec. 3, 2003

When the Outlook plugin filters mails, it classifies them as either spam or potential spam, and can put them in appropriate folders. In the spam/potential spam folders, there is a "Recover from Spam" button available, and in other folders there is a "Delete as spam" button. These buttons add the message to the training database as well as taking the appropriate action. One thing I don't see, however, is a means of confirming the classifier's decisions as correct. A "yes, that is spam" button for the spam folder, and a "yes, that's ham" button in non-spam folders. As I'm starting from a very small message base, I worry that correct classifications are still somewhat based on "luck", and training based on correct decisions would help to increase both my and the classifier's confidence level. I can do this by regular retraining, but that has 2 disadvantages: it's much clumsier than simply clicking on a "clever boy!" button, and it relies on me not deleting messages until I do a training run. Much of the ham I get is "read and forget", so I'd rather delete immediately. When I get a chance to dive into the code, I'll see how hard this would be to implement. Paul.

7 8

RE: [Spambayes] Outlook plugin - training
by Piers Haken Dec. 3, 2003

Dec. 3, 2003

I don't believe you need this. I think that the classifier automatically trains on messages as they arrive (or at least on messages that it's sure about). You only need to retrain if it has made a mistake, or if it's unsure. Piers. > -----Original Message----- > From: Moore, Paul [mailto:Paul.Moore@atosorigin.com] > Sent: Wednesday, November 06, 2002 2:09 AM > To: Spambayes (E-mail) > Subject: [Spambayes] Outlook plugin - training > > > When the Outlook plugin filters mails, it classifies them as > either spam or potential spam, and can put them in > appropriate folders. > > In the spam/potential spam folders, there is a "Recover from > Spam" button available, and in other folders there is a > "Delete as spam" button. These buttons add the message to the > training database as well as taking the appropriate action. > > One thing I don't see, however, is a means of confirming the > classifier's decisions as correct. A "yes, that is spam" > button for the spam folder, and a "yes, that's ham" button in > non-spam folders. > > As I'm starting from a very small message base, I worry that > correct classifications are still somewhat based on "luck", > and training based on correct decisions would help to > increase both my and the classifier's confidence level. > > I can do this by regular retraining, but that has 2 > disadvantages: it's much clumsier than simply clicking on a > "clever boy!" button, and it relies on me not deleting > messages until I do a training run. Much of the ham I get is > "read and forget", so I'd rather delete immediately. > > When I get a chance to dive into the code, I'll see how hard > this would be to implement. > > Paul. > > _______________________________________________ > Spambayes mailing list > Spambayes(a)python.org http://mail.python.org/mailman/listinfo/spambayes >

10 18

Advantages of S?
by Gary Robinson April 26, 2003

April 26, 2003

At present, it appears that S may perform better than relatively untweaked graham (as I guess bogofilter is using), but about the same as tweaked graham as spambayes is using, but has the advantage of showing a spread of results. As I see it there are two possible advantages to the spread: 1) emails in the middle could be classified as "you should manually check these", or 2) the user could adjust for the level of sensitivity he wants (similar to what happens now in Entourage which is otherwise brain-dead as a spam filter). (1) is a attractive but one thing worries me: it may give people a false sense of security -- they may think they only have to look at the middle ones, but occasionally even extreme cases of seeming to be spam might be legit. On the other hand, on those rare occasions, the sender can always try again if he doesn't get a response! ;) Just a couple of thoughts. --Gary -- Gary Robinson CEO Transpose, LLC grobinson(a)transpose.com 207-942-3463 http://www.emergentmusic.com http://radio.weblogs.com/0101454

8 19

Two Stage Plan
by Gary Robinson Jan. 16, 2003

Jan. 16, 2003

I want to suggest a two-stage plan to solve the spam problem. I'm not sure if it makes sense, but it's interesting enough to me that I decided to share it to see what other people think. FIRST STAGE Many of you are aware of the http://wecanstopspam.org idea, whereby: --If a lot of real people use it in the sigs of real emails, spam filters will get trained to see it as a very strong indicator of being legitimate. Thus, it will have become a sort of "virtual whitelist". I see this as being able to counteract, to some extent, the fact that spammers will be trying to use words with no very spammy associations. Instead, this technique puts the stress on "hammy" words, in particular this very hammy indicator. --If the URL does become widely used and is accepted by filters, of course spammers will want to include it too. But at that point, it will be popular enough that filter authors will be motivated to make sure that only visible, clickable versions of the URL are given a high hamminess value. So spammers would have to, in effect, advertise the wecanstopspam.org website and provide a convenient link. --The URL would contain information about how to combat spam, as it does now, but hopefully much better written and presented, as the site evolves under community guidance. So spammers that include it will be helping their targets to fight spam. SECOND STAGE The problem with all possibly foolproof anti-spam approaches, such as the pay-to-spam approach, or the camram one (http://www.camram.org/), is that there is a huge chicken-or-egg problem. The world really has to settle on one solution and get a real critical mass of users in order for it to work. Now, if in fact it gets to the point that spammers are sending the http://wecanstopspam.org URL to millions of users a day (or even if it doesn't, but millions of individuals are using it because the virtual whitelist aspect), then there will be enormous power associated with the wecanstopspam.org site. That is, that site may then, all by itself, have the power to determine what the world standard solution is by announcing it on the site. What will it be? That would be determined by some sort of community process. Maybe online voting, or maybe a conference where people would discuss and finally vote on the solution. CONCLUSION If: --A compelling enough meme could be crafted that people would want to include the URL in their sigs so that it would spread in a p2p viral fashion, and --It is in fact possible for filters to only give credit to the token when it is visible and clickable, then it seems to me that this could serve as a realistic means for solving the chicken-and-egg problem, thereby creating a single dominant standard with enough critical mass to actually work. The basis for it is that it avoids the chicken-or-egg problem in the first stage by leveraging existing spam technology. It can do that because the substrate is already in place for the idea to get to critical mass, in the form of existing adaptive spam filters such as Graham's. Then when it gets to critical mass, spammers will want to co-opt the token, except that in the act of doing that they give the wecanstopspam,org site enough power to enable the world to agree on a foolproof solution. Now, I realize the above may be crazy since I haven't thought about it for that long. But I just thought it was perhaps interesting enough to be worth sharing. Feedback? --Gary -- Help your email get through while making life harder for spammers: use http://wecanstopspam.org in your sig. Gary Robinson CEO Transpose, LLC grobinson(a)transpose.com 207-942-3463 http://www.transpose.com http://radio.weblogs.com/0101454 >

4 3

bayesian research
by alan Dec. 31, 2002

Dec. 31, 2002

Hi, I have been given the task of reseaching Bayesian mail filters for my final year Univeristy dissertaion. I have been finding brick walls at every turn. I know paul graham is great, but just about every one talk about his plan for spam, but i need a start place. I have set my system to allow relays and have 1000's of spam examples. Any ideas where i should start? thanks for your time Alan Paice Cambridge, UK

5 4

Mail classifiers, training sets and technical docs
by Niek Bergboer Dec. 30, 2002

Dec. 30, 2002

Hello, Like many others, I suffer from Spam, and while surfing the web I came across your Bayesian mail classifier. However, since I am also doing my PhD research in the field of machine learning, I am especially interested. Specifically, I am using machine learning techniques (including classifiers) on images, but the application to email seems interesting as well. First off, I tried to find some in-depth technical documentation about your system, but I was unable to find it. Could you direct me to any literature references or papers on which the work is based? Being involved in machine learning, there are of course a number of "standard" questions that immediately pop up: Does the SpamBayes framework use any training before it gets shipped to the user? That is, does the user start out with a completely "dumb" system for which he has to provide _every_ single spam/ham example, or does the system come with a "basic training set" so that the system has some classification capabilities even before the user has specified any examples. If so, what kind of training set do you use? How large is it, and what is the dimension of your feature space? And do you plan to make a large training set available? In addition, I was wondering about the kind of classifiers that could be used. It seems to me that SpamBayes basically is a binary classifier: mail has to be classified as either ham ("1") or spam ("0") (or vice versa, if you like that better). In addition to (naive) Bayesian classifiers, there of course exist more. For example, a classifier that has been around for a while, but has only just begun to be viable (do to new training techniques) is the Support Vector Machine (SVM). In its basic form, this is a binary classifier (though multi-class problems can be handled as well nowadays). Theoretically, one could of course also use a very simple and crude (k-) Nearest Neighbor classifier, though one would need a large training set for this to work well. Based on which criteria was the choice for using a Bayesian classifier made? I wish you the best of luck and success with the project. Good work! TIA, Niek -- N.H. Bergboer - n.bergboer(a)cs.unimaas.nl University of Maastricht - +31-43-3883901 Institute of Knowledge and Agent Technology "I have been asked, 'Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?' I am not able to rightly apprehend the kind of confusion of ideas that could provoke such a question." Charles Babbage

4 4

RE: [Spambayes] Mail classifiers, training sets and technical doc s
by Diffenderfer, Randy Dec. 30, 2002

Dec. 30, 2002

> In addition to (naive) Bayesian >classifiers, there of course exist more. For example, a classifier that >has been around for a while, but has only just begun to be viable (do to >new training techniques) is the Support Vector Machine (SVM). In its >basic form, this is a binary classifier (though multi-class problems can >be handled as well nowadays). Theoretically, one could of course also >use a very simple and crude (k-) Nearest Neighbor classifier, though one >would need a large training set for this to work well. IIRC the words "Support Vector Machine" appear in some patent that Microsoft has on classification technology. Reference to it is in the spambayes archives somewhere. :-) Needless to say, building a system based upon something that M$ has a patented interest in is probably a losing proposition.

1 0

Re: [Spambayes] Missing 2.0 compatibility lines in storage.py
by François Granger Dec. 29, 2002

Dec. 29, 2002

At 00:41 +0100 on 29/12/2002, in message [Spambayes] Missing 2.0 compatibility lines in storage., François Granger wrote: >Just added these cutted and pasted lines at the biggining of the file: > >try: > True, False, bool >except NameError: > # Maintain compatibility with Python 2.2 > True, False = 1, 0 > def bool(val): > return not not val Same for Coprus and Hammiebulk. -- Recently using MacOSX.......

1 0

Missing 2.0 compatibility lines in storage.py
by François Granger Dec. 28, 2002

Dec. 28, 2002

Fresh install on MacOSX with the standard Python distribution: [fbg:/volumes/OS99/spambayes] fgranger% python Python 2.2 (#1, 07/14/02, 23:25:09) [GCC Apple cpp-precomp 6.14] on darwin Type "help", "copyright", "credits" or "license" for more information. Fresh download from CVS, I get the following errors. [fbg:/volumes/OS99/spambayes] fgranger% python hammiebulk.py Traceback (most recent call last): File "hammiebulk.py", line 53, in ? import hammie File "hammie.py", line 5, in ? import storage File "storage.py", line 57, in ? NO_UPDATEPROBS = False # Probabilities will not be autoupdated with training NameError: name 'False' is not defined [fbg:/volumes/OS99/spambayes] fgranger% python hammie.py Traceback (most recent call last): File "hammie.py", line 5, in ? import storage File "storage.py", line 57, in ? NO_UPDATEPROBS = False # Probabilities will not be autoupdated with training NameError: name 'False' is not defined [fbg:/volumes/OS99/spambayes] fgranger% Just added these cutted and pasted lines at the biggining of the file: try: True, False, bool except NameError: # Maintain compatibility with Python 2.2 True, False = 1, 0 def bool(val): return not not val -- Recently using MacOSX.......

1 0

Spambayes Training.
by Tim Uckun Dec. 23, 2002

Dec. 23, 2002

Hello, you want some unsolicited advice? Considering that your spam filter will always encounter all incoming mail you could perhaps build in the training to your filter. I imagine a scenario like this. I get some spam your filter did not catch. I then forward that email to .. a) Myself (more on this) b) DetectedSpam(a)mydomain.com If I forward the email to myself I add something to the first line like --HeySpamBayesThisIsSpam-- If your filter catches this on the first line of an email it then learns from it. Similar things can be done with ham. This way you can even start with a completely blank database and then train your filter perfectly. The beauty is that you don't need some special mail client and it makes no difference whether spambayes is running as a proxy or through procmail. ---------------------------------------------- Tim Uckun Mobile Intelligence Unit. ---------------------------------------------- "There are some who call me TIM?" ----------------------------------------------

3 3