McAfee Granted Far-Reaching Spam-Control Patent

Titusdot Groan writes "Infoworld is reporting that Network Associates, makers of McAfee, have been granted a broad anti-spam patent. The patent covers "compound filters, paragraph hashing, and Bayes rules" and was filed in December of 2002. The patent appears to affect Spam Assassin, Spam Bayes and many other anti-spam products and services. As an aside Paul Graham's "A Plan for Spam" was published August 2002."

Comments

And so what?

[Anonymous Coward]

Some legal experts have expressed doubt that the patent will stand up to challenges given its wide scope.

I think Spam Bayes/Assasin can rest easy to be honest.

Re:Prior usage?

[Iamnoone]

why the "logo" for this article on Slashdot is "fork, knife and spoon"

Because they are well known, common items and they have the word patented stamped on them - trying to point out the problem with patents for things that are "obvious to those versed in the art".

This won't stand for long

[speters]

Mr. Graham's article discussed the Bayesian filter he was already using in the article. Several other spam filters, including SpamAssassin, added Bayesian processing shortly after the article was published. The patent holders would have a very hard time trying to go after those who had Bayesian rules processing prior to the patent being filed.

Re:This won't stand for long

[shlomo]

Actually this was first proposed by Sahami et al in 1998! Graham never did his hw either

"A bayesian approach to filtering junk e-mail"

In AAAI-98 Workshop on Learning for Text Categorization, 1998.

Patent Link

[tronicum]

Here is a link to the patent at USPTO

They are refering to Bayes a lot, beside really simple stuff like "hashing an paragraph with MD5". That is like making an patent on the progress bar (that exits too!)

It is only for US anyway....

Bad McAfee

[sufehmi]

OK, McAfee is officially in my "bad company" list now.

What are they thinking exactly by patenting Bayes rules, etc ? So take the best from open-source community, and then patent them under your own name, eh ?

I'll share some info about McAfee now:

• For a better antivirus, use NOD32 instead (they never missed a single virus in 6 consecutive years).
  
• For better anti-spam software, use POPfile instead (and it's free)
  
• For anti-spyware, use Spybot instead (and it's free)
  
• For firewall, use ZoneAlarm instead (and it can be free)
  

Do I miss anything ?

I think we should distance ourselves to nasty companies like this. Let's speak with our money.

procmail as prio art?

[richieb]

Earliest mail filtering I remember is procmail, which is still alive and well.

It was first published in 1990.

Re:procmail as prio art?

[ram4]

Larry Wall's "mailagent" initial implementation appeared in comp.sources.unix in 1989 IIRC, in his "dist" package (the package that contains metaconfig, the Configure script generator).

This early implementation was sending acks to people and was also processing "commands" embedded in the message. I think it pre-dates procmail.

The "mailagent" script finally grew outside of the "dist" package and became a standalone mail filter, with a nicer syntax than procmail. Search for it in CPAN if you want to compare.

Re:Prior usage?

[fw3]

No.

What matters is date of invention, not filing.

If NAI can demonstrate that they were working on a Bayes approach prior to Graham's work, then they may indeed have thougth up the idea first.

<researching> ....

Clearly Graham was not the first to think of Bayes as an approach to spam This paper ca 2000 predates both and is cited by the USPTO in the patent as a reference.

Contrary to prevailing /. wisdom the patent process does actually involve research on both the Patent office and the applicants part.

Re:Death by patents and spam?

[Doctor O]

Yes he does. It's why he wrote "Let's not forget that some of the most useful new uses for e-mail are webbased; that the underlying technology is SMTP doesn't really matter to most people."

It doesn't matter whether that HTML form uses SMTP or whatever, it's the same form to the user. So for those, E-Mail can be easily replaced with something else.

UGH, It Gets Even Worse for NAI

[sabat]

The following text is currently on SpamAssassin's site (see http://spamassassin.org/prehistory/) -- keep in mind that it's the basis for the patent. Emphasis below is mine.

SpamAssassin Prehistory: filter.plx

Before there was SpamAssassin, there was Mark Jeftovic's filter.plx. This was a 'context/keyword spam filter', which came up with the basic scheme of what we use in SpamAssassin: namely, named rules identifying spam-like 'features' of the mail, each rule has a score, and once the number of 'strikes' goes above a certain threshold, the mail is marked as spam. And written in perl, of course ;)

I (Justin Mason) used this for several years, adding a few small snippets of code; eventually though, the code was getting a bit stale, and Mark seemed busy on other stuff, and I had a few thoughts about some improvements I could do with a total rewrite ;) -- so I decided to recode from scratch under an open-source license, and that was SpamAssassin.

Unfortunately the original site at http://antispam.schmooze.net/filter/ is no longer up, but the Internet Archive has a snapshot of it from December 1998 here.

Also courtesy of the Internet Archive, the change log of filter.plx is here, spanning June 1998 to August 1997.

Finally, Mark was kind enough to dig up a source code tarball for filter.pl-1.02d.tar.gz (20k). This is the 1.02d release, February 1998.

Whatever you do, don't actually run the code -- spam nowadays looks nothing like spam did back then, before e-mail clients grokked HTML. Plus I don't think Mark wants to get bug reports at this stage, it's been 5 years ;) This page is here instead to document the history of this project.

--j. Jul 14 2003 jm

This patent is not a problem.

[Anonymous Coward]

You just need to read the claims...

Every one of the claims either includes the phrase

"wherein the paragraph hashing excludes at least one of a first paragraph and a
last paragraph of content of the electronic mail messages"

or is dependent on a claim that includes that phrase.

In short, as long as you don't clip off the first or last paragraph, the patent does not apply.

A good (in the FSF sense) side effect of this patent is that the methods given _absent_ the clipping (or, for that matter, absent any of the other processing bits listed in the claims) ARE NOW PUBLIC DOMAIN AND CANNOT BE PATENTED, EVEN BY MCAFFEE! :)

So, just don't worry about it, don't clip either the first or last paragraph, and you're fine.

You can also dissect the patent on the grounds that it only covers spam filters that both hash paragraphs AND do bayesian filtering... not either one alone ( the so-called "three-legged stool" rule in patent law - given two patents, one patent claiming only a subset of the stuff in the other patent, then the smaller patent wins)... and in this case, IF this is the smallest patent McAffee can make, we're in good shape.

(disclaimer, I am not a lawyer. But I do occasionally have to do patent stuff for work. And I'm also the prime author of CRM114, which is a moderately hot-shot spam filter among other things. So I'm not entirely third-party disinterested. But I'll also say that I'm not worried. :) )

Speaking as the author of POPFile I say...

[JohnGrahamCumming]

Fuck that(*)

John

(*) McAfee Legal: "Fuck that" is a technical term that indicates that significant prior art exists that invalidates a patent.

Sounds like another job for PUBPAT or EFF!!

[the_rajah]

These improper patents have to be fought by someone. See PubPat and, if you agree with what they are doing, make a contribution. The Electronic Freedom Foundation is doing some of this worthwhile work, too.

"Do the Right Thing. It will gratify some people and astound the rest." - Mark Twain

Re:A good example for EU

[flossie]

Yes. There are three pillars of government to the EU - the Parliament, the Commission and the Council of Ministers. Legislation requires agreement from all three.

The European Commission (appointed by ministers in each state) proposed to introduce software patents along US lines. The European Parliament (the only democratically elected pillar) amended the proposed directive to prevent software patents. The Council of Ministers (made up of ministers of each state (the ones who appoint the Commission)) have just voted for an even more extreme pro-patent position than the original Commission proposal. In order to defeat this, an absolute majority* of all the Parliament's MEPs must vote against the Council's new version.

* this is an absolute majority of all elected MEPs, not just a majority of those who bother to vote.

Prior Art

[epsalon]

Bayesian spam filtering was invented and implemented long before Paul Graham's "A plan for spam". A project I wrote for a course in July 2001 that does Bayesian filtering was based on papers suggesting to do the same for spam, the predated that time. I used the same technique in 2001 to write an internal ad filter (something to filter ads inside mailing list postings) using bayesian methods, so this is clearly prior art.

Re:Invalid stupid patent.

[cyberlync]

Patent fees. Getting a patent is not a free service, in fact its quite expensive. Its gotten to the point that an individual can't even begin to patent anything unless he is wealthy or has a large corporate backer. Without spending much on research the USPTO become quite profitable for the government.

Disputing Patent

[Nuclear Elephant]

If anyone's research was published before this patent was filed (Dec 2002), you can request an investigation by submitting a letter and a copy of your research (and of course, when it was published) to:

US Patent Office
PO BOX 1450
Alexandria VA 22313

Make sure you reference Patent 6,732,157

I was told it would be routed to the right department. The patent was filed in December 2002, and I know much of everyone's research in the anti-spam arena was published long before that.

Re:Prior usage?

[sapped]

"fork, knife and spoon" (in German we call it "Besteck" but I know that the english language has no equivalent for it)?

The english equivalent of Besteck is cutlery.

In the USA they refer to it as silverware. (Even if it is made out of white plastic.)

Stop complaining, start proving

[CaptainFrito]

It's easy. Just make the prior art available on a website somewhere, with verifiable dates. Use it, one-by-one, to demonstrate that their solutions were obvious to one skilled in the art at or before the time they filed, right down the claims list. Once the information is public they won't ever try to enforce it, because patents -- invalid or not -- are considered valuable assets, particularly if the company has the financial resources to fund a lawsuit to enforce. So they'll prefer to simply let sleeping dogs lie. And if they do try to enforce it, the defense will be straightforward an inexpensive.

IANApatentL but I believe that patents only prevent others from commercializing a claimed invention. OSS is not a commercialization per se. It is simply a public disclosure of a particular method. Patents too are a public disclosure of particular solutions. But it would be hard to prove [to me, anyway] that a patent assignee would be damaged by publicly telling others something they had already publicly disclosed themselves.

As far as I am aware, time-limited monopolies are permitted in exchange for the disclosure. It it completely legal and non-actionable, as far as I understand the law, to use the disclosure for personal non-profit use. So, for an anti-spam filter, it would be hard to argue that anyone, even businesses would be using such a system for profit. I would argue that it is a necessity under law to prevent things in the workplace that harass or oppress or [truly] offend, which a lot of spam does. So, rather than for profit it is a legal necessity (in many countries, certainly in the jurisdictions that would be asked to enforce this patent).

Complaining that the world is unfair -- a fact known to most adults -- is profoundly useless.

Re:Invalid stupid patent.

[j-turkey]

Patent fees. Getting a patent is not a free service, in fact its quite expensive.

Indeed, it appears that there is some money to be made from patent applications. Upon some quick research, a patent application is quite expensive...and it's clear that the USPTO is not spending that money researching prior art. However, to shut the anti-corporate camp up a little (and Cyberlync, this is not a rip into you by any stretch)...if you read the fee schedule, there are separate fees for a small entity and "other than a small entity". The latter half have to pay double for everything.

Prior Art

[tacocat]

bogofilter was released as follows: from their web page

Revision 1.1 - (download), view (text) (markup) (annotate) - [select for diffs]
Sat Sep 14 22:15:20 2002 UTC (20 months, 2 weeks ago) by adrian_otto
Branch point for: bogofilter-vendor
Initial revision

Prior Art = Mozilla ?

[darinf]

Bayesian spam filtering was first added to the Mozilla codebase on September 9, 2002 by Dan Mosedale. The first major release to include it was Mozilla 1.3 Alpha, which was published on December 13, 2002. Is that not prior art?

I have prior art!

[jefferson]

In 1999 I wrote a paper comparing Naive-Bayesian classifier learning to a good rule-learning algorithm for classifying email, including spam. And showed how well the Bayesian classifier worked on spam. The paper is linked from Paul Graham's spam research page.

Here is the reference:

Provost, J (1999). Naive-Bayes vs. Rule-Learning in Classification of Email. The University of Texas at Austin, Artificial Intelligence Lab. Technical Report AI-TR-99-284

I should mention that I don't think I'm the first to use Naive-Bayes on email. I think some folks from Microsoft did it in 1998, and there may be others too.

Re:I worked at McAfee...

[belphegore]

I worked at McAfee too back when it was Network Associates, after they bought Deersoft, which I founded. Deersoft you'll recall was the company that made and marketed commercial versions of SpamAssassin. I'd just like to point out that we, the Deersoft folks, had nothing whatever to do with this patent. It appears to have originated with the prior-to-Deersoft SpamKiller product (the windows desktop app).

Also, the open source SpamAssassin project is likely 100% in the clear on this patent, even if it is valid (which is a separate question), since Network Associates and all of its employees who worked on the open source project have filed CLAs or CCLAs with the Apache Foundation. Section 4 of each document is worth a read. Looks like anyone who licenses a copy of the SpamAssassin code from ASF gets the right to use Network Associate's patent. Though IANAL.

Microsoft has at least one patent in this area.

[crisco]

Dating back to December of 2000, Patent # 6,161,130 appears to cover some of the same techniques.

Abstract

A technique, specifically a method and apparatus that implements the method, which through a probabilistic classifier (370) and, for a given recipient, detects electronic mail (e-mail) messages, in an incoming message stream, which that recipient is likely to consider "junk". Specifically, the invention discriminates message content for that recipient, through a probabilistic classifier (e.g., a support vector machine) trained on prior content classifications. Through a resulting quantitative probability measure, i.e., an output confidence level, produced by the classifier for each message and subsequently compared against a predefined threshold, that message is classified as either, e.g., spam or legitimate mail, and, e.g., then stored in a corresponding folder (223, 227) for subsequent retrieval by and display to the recipient. Based on the probability measure, the message can alternatively be classified into one of a number of different folders, depicted in a pre-defined visually distinctive manner or simply discarded in its entirety.

John Graham-Cumming (author of POPFile) has this to say about Microsoft's patent:

1. POPFile was not designed for the sorting of spam from legitimate mail it is a general email classification system.

2. I believe the patent to be invalid because of the ifile system being prior art. ifile dates back to at least 1996 while the patent has the date June 23, 1998 on it. The patent does not acknowledge ifile's existence. Evidence of ifile being prior art can be found in the ifile change log http://www.nongnu.org/ifile/ChangeLog and the original README (http://www.nongnu.org/ifile/old/README-0.1A) which shows the date: Released Sat Aug 3 20:49:01 EDT 1996

3. If Microsoft were to sue me and win I would be happy to pay them every penny that I have made from POPFile ($0.00) :-)

Some very early prior art

[nero4wolfe]

There were open source spam filtering systems using Bayesian rules as early as 1996. Look at the url http://www.nongnu.org/ifile/README-0.1A

I played with ifile at one time...

Previous Works

[cryonv]

Hmmm...

I distinctly remember utilizing VirusX on the Amiga in the EARLY 90s. Along with many other virus tools of the 80s...