Slashdot Mirror
Cantametrix Plans To Track All MP3s On The Web
Akilesh Rajan writes: "A Stereophile article reports that Cantametrix is further developing its MusicDNA system for identifying and tracking all MP3s on the Internet.
MusicDNA's use of DSP (Digital Signal Processing) technology and psychoacoustic modeling allows it to analyze an MP3 and immediately tell what song it is, and so also recognize who, if anyone, owns its copyright. Company reps explain one possible application: 'A MusicDNA Analyzer can be located, for example, on the Web crawler of a large search engine, to ensure that the search engine only points to legal music.'" I could see this working a lot better if all the music on the Web was pristine and complete -- which it's not.
Nothing in this technology _stops_ it from being usable in the last sense- a way to quickly be pointed to the rest of an artist's freely available (typically low bit rate) catalog. That, not 'monetization', is the new concept: the idea that for the first time a good but poorly resourced artist would have the same information distribution resources as the majors- the majors fight and spend billions to try and get some produced 'artist's name into peoples' ears, so that the consumer knows what they're hearing and where to buy more (at your local CD store, of course). For the first time this might be truly decentralised so that anyone, anywhere, who was listening to some anonymous and obscure song they liked, would be able to get the information. So it's on the radio? Hold up a mike and tape the radio. Pirate radio? Same deal. Old cassette tape that never had a label? No problem. mp3 marked "Metallica-One.mp3" erroneously? No problem...
At that point, you start having a free market again- at that point good local or indie bands or musicians, or really specialised musicians (noise, trad jazz, ragtime, serial composition) can begin shortcircuiting the lines of distribution and undercutting the majors by the simple expedient of 'who cares if I can't make money at this, nobody does but _I_ can afford to voluntarily give out mp3s etc. and FIND MY AUDIENCE'. At a stroke, the barriers to entry for an entire industry fall, and genres like jazz can survive (contrast this with at the major labels, which not only will not support jazz but are known to actually destroy irreplaceable master tapes to save storage costs- refusing to allow anyone to salvage the masters).
I really hope these people have enough sense to become this type of general resource. The risk is that the majors will not permit information to be stored for any music other than major label 'protected' music, and so the more obscure or indie stuff will turn up as 'no matches'.
couldn't they send out threatening emails to anybody who has an MP3 with (for instance) Metallica in the file name, with roughly the same effect?
They did that already. Didn't work all that well (because lots of things other than Metallica files have that in the name, and lots of Metallica MP3s aren't named so simply). It also caused a firestorm.
Distinguishing between what is real and what's not is probably only useful in court... (correct me if I'm wrong...)
Whether it's a real Metallica song or an unlicensed cover of it doesn't matter. They're both copyright violations.
If the technology really does work - or even works moderately well as a bird-dog - using it for a webcrawler to hunt for infringers may work - and be within existing law. They can check manually before going to court. If it's good enough, they can weasel-word a cease-and-desist order and not get much problem from occasionally sending one to a host of a misidentified file.
No law changes required. If they finally buy a clue and go after the >hostsindexers, they'd be on completely solid legal ground, and the litigation would be reduced to:
- Did the defendant knowingly host a copyrighted work without obtaining the proper license?
- Did he refuse to take it down in response to the cease-and-desist order?
- Does the plantiff hold the copyright (or otherwise have standing)?
- (In the first few cases) is such hosting fair use?
But even if it's NEARLY perfect it will sometimes misidentify a non-infringing work. If it does this even once, it opens any subscription hosting service that uses it to civil action for contract violation by its customers.
As we've seen, in free competition the indexing services that use a filter will lose to those that don't - because they'll lose the portion of the customer base that doesn't care whether they're downloading a copyrighted work. And any flase-positive flakeyness in the technology would produce the same sort of flap as the nannyware web filters. This should preclude attempts to pass and enforce a legal mandate, on first amendment grounds.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
If they finally buy a clue and go after the >hostsindexers, they'd be on completely solid legal ground ...
...
Make that:
If they finally buy a clue and go after the hosts rather than the indexers, they'd be on completely solid legal ground
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
The same law that prevents you from putting the mp3's on the server in the first place? :-)
That would correct wrongly named and labed files and make people's collections look neater.
--
2 -- an expert system powerful enough to comprehend and categorize musical information, that could tell a licensed recording of Mozart from a bootleg NIN concert, i.e. practically full-blown Artificial Intelligence.
- Convert the
.ogg, .mp3, etc. to .wav or some other transparent linear PCM format.
- Use a low-pass filter to go down to the low frequencies where the bass line lives.
- Use a similar algorithm to the "beat finder" in many XMMS/Winamp plugins, along with the Fourier transform, to reduce the wave to a list of notes being played.
- Taking into account transpositions and speed changes, pattern match with MIDI files from the ASCAP, SESAC, BMI, and RIAA libraries.
- Sue.
This technique also would have caught "Ice Ice Baby" (really "Under Pressure") and "Come As You Are" (really "Eighties").Will I retire or break 10K?
One possible workaround: prepend a random barely-audible rumble to each MP3.
--
Yeah, right.
I'd like to see them do this, and encompass the myriad of different protocols and formats that abound on the web today, plus the ones that will be designed just to break it.
I think that simple passwords, encryption, steganography, and file-sharing will each be enough to defeat this, but who knows, maybe we'll have to go to something really sophisticated, like trading over IRC, or ratioed ftp...
Companies that base their business model over scare tactics just crack me up...
---
pb Reply or e-mail; don't vaguely moderate.
pb Reply or e-mail; don't vaguely moderate.
Ohhh, maybe they will sell this technology too. Then I can finally categorize all my mp3's. I've got a couple that I can't tell whether they're Christina Aguilera or Britney Spears. With this cool technology I could just run it and it'd go grab some info from cddb.com and voila. This rules! heh. ;-)
Time to convert *.mp3 to *.ogg.
and they don't discover serious flaws while doing it (say, discover 100 different tracks that have the same fingerprint)
More likely to end up with 100 different fingerprints for the same track. Well the same track so far as humans are concerned, different tracks because of the way they have been ripped and processed. Remember that MP3 uses a lossy compression.
But after a while, the pirates get greedy and form larger clumps, which makes them more visible to the authorities. Eventually the clump gets raided and everyone scatters. Some form small islands again and grow over time and the cycle continues.
As for me, my particular island got raided and somehow I've never gotten back into it. Other than the casual Napster use, which doesn't count because Napster is an island that was allowed unmitigated growth for long enough via unanswered legal questions that it grew to an immense size and its members are now powerful enough to openly do damage to the authorities (eg. Metallica vs. The Fans) and maybe even force The Rules Of The Game to change.
That said, pirate islands will still be able to play by whatever rules they want.
--
So this means that a search engine is now going to need to download every mp3 file it finds each time it crawls the net? Boy I feel sorry for mp3.com when one day some machine @ inktomi decides to pull down *every* mp3 on the site. Thats gonna be an expensive bill...
What it will do is create a new genre of music, "Sonographic Clone Rock." Creating a program that can identify sonic patterns across encoding formats, bit rate qualities and whatever slight effects can be added to a copyrighted song will make it broad enough set off false alarms with something as simple as a spoof song.
Or at least that's my prediction.
I could see this working a lot better if all the music on the Web was pristine and complete -- which it's not.
Now, that's funny. I could see this working a lot better if all the music on the radio was new and original. How would you tell one Britney Spears song from another, or from any Ace Of Base title?
© Copyright 2000 Kristian Köhntopp
oh i thought you were talking about SDMI...
Why does everyone forget that:
a) they don't have to send out a cease and desist letter directly from the output of the program. Lawyer for RIAA will use the output of the program to find where an infringing MP3 might be, use that to go look for him/herself, and then decide if the C&D letter is appropriate. The program will just be used to improve lawyer efficiency.
b) they don't have to catch every bootleg MP3. Just enough to put a chill on the free speech issue.
c) they only have to search for a subset of their copyrighted songs. Those top 100 that are currently popular. They aren't loosing much money on the rest, and those are scarce on the Net anyway.
d) the RIAA et.al., can run their own damn search engines on mainframes with all the money they've ripped from artist.
Hasn't Microsoft proven time and again that software technology doesn't have to be good to fool the sheople, just good enough?
Aah, change is good. -- Rafiki
Yeah, but it ain't easy. -- Simba
This and that stupid map of the Internet that was on /. the other day are more amusing than not. I have gigs of MP3's and various other files that are probably questionable and I certainly haven't seen anyone that shouldn't be there in my iplogs scanning those files. People scp the files from me all the time so it does make me wonder what exactly they are tracking.
At what price learning? At what cost wisdom? The price is a man's peace of mind, and the cost is his life.
By basing the pattern matching strictly on low-range frequencies, you're FFT/"beat finder" algo isn't going to catch any of the patterns in the higher frequencies (thus being b0rked by songs with sounds strictly >= 1000Hz or so, or different songs that use the same base sounds, like, say, every rap song in existence). Further, you're planning on running this algorithm (which requires doing a digital format change and computing a FFT, neither of which is cheap in terms of disk or CPU) on every song retrieved by a search engine (more resources in terms of bandwidth)? The search engnies would laugh at you if you proposed they spend money on this to actually put it into service on their machines. Oh, and now any search with music terms in it takes a leisurely 12 hours to complete if you match more than about 3 songs. Also, given the inherently distorted nature of the found song once you've bandpassed it, wouldn't you have to do the same thing to the MIDI files in the auth lib?
So even if your non-AI algo were to work reliably (which I highly doubt), it would be prohibitively expensive in terms of system resources (now or a decade from now).
--
News for Geeks in Austin, TX
A privacy amendment will also us to quote it like we do now, such as, "take the 5th", "1st amendment rights", So we need an amendment that gives the people basic privacy rights, that pertains to the 21st century, and while were in there, we could probably solve some copyright use issues as well."
Contrary to popular belief, I don't actually make my website for other people to look at.
I remember not long after I got an internet connection (through the U, august of 94), this big brouhaha happened about some people (Unisys? Lawyers acting for them? How quickly brain cells die when soaked with hard alcohol...) that were supposedly releasing a worm onto the Internet to "ferret out" patent-infringing GIFs...
The small problem with that was, it was impossible. Even if some secret header code existed in "licensed" gifs, which to my semi-sketchy knowledge about graphics file formats does not (unless maybe gifs from "licensed" authoring tools had some sort of characteristic fingerprint like "made by gimp" or whatever), imagine for a second the difficulty of finding, cataloging, and determining the ownership of every gif on the net.
Now take all of the previous difficulties of this type of InfringeWare, undiminished and in fact probably heightened, and add to them the fact that now instead of being concerned about the file format (a relatively fixed thing), you're trying to judge infringe/not-infringe based on the content itself. This would require one of two things to work (from what I can tell talking out my ass on slashdot @ 5am whilst drinking):
- 1 -- a complete DB of every song ever recorded in any digital format, possibly at different bitrates. This would be for a "dumb" approach using pattern matching/comparison (like a global regular expression for mp3 contents, which might actually be a nifty hack for a local program ("computer, play me someting hardcore, with lots of drums" and the machine looks for patterns in the files themselves to spit something out into the soundstream)).
- 2 -- an expert system powerful enough to comprehend and categorize musical information, that could tell a licensed recording of Mozart from a bootleg NIN concert, i.e. practically full-blown Artificial Intelligence.
Both of these things seem less than likely to occur anytime soon. (If they had the former they'd be whoring it out a la terraserver's approach to space imagery, if they had the latter I hope as a human being that we could find more meaningful things to do with true AI than searching for mp3z!No, I think that this is just hot air intended to scare people into thinking the Big Bad Patent/Copyright-Holding Wolf is Just Around The Corner, so It's Time To Shape Up And Quit Trading Mp3s You Little Monsters... Another option is this is a vaporware company trying to feed of the greed and stupidity of the record labels...
--
News for Geeks in Austin, TX
Let's say it can only match perfect rips (ok, of course it will have some sort of tolerance, but anyway), then this would be useful to weed out bad copies. How many times have you downloaded an mp3 to find that it has been encoded really bad, or worse still recorded off the radio, complete with back-announcements? Very, very annoying, because they never manage to announce the next song right.
The result was a piece of music, a performance, that had never existed before, done are a tempo that had more punch and groove.
This worked out really well. But now I have a bit of music that is something the original artist never recorded.
Who owns the copyright to that, and how would it sort out according to this proposed technology?
"It is a greater offense to steal men's labor, than their clothes"
The other problem is, will they adhere to robots.txt files? If they do, then bypassing the mp3 'sniffer' is a joke; if not, then they should be considered to be violating the explicit denial of a site to allow 'hacking tools' such as a search bot and are still in the wrong. In other words, this will either be uneffictive, or treading illegal water territories (and not necessarily in the vein of copyright infringement).
"Pinky, you've left the lens cap of your mind on again." - P&TB
"I can see my house from here!" - ST:
From the I can't believe they're this stupid dept.
Could this be the Technolibertarian's dream come true and the end for constant vigilance and street corner phophetizing as we know it? FuckedFromtheOutset has announced a preliminary effort to start the planning process on some more vaporwear. Music DNA, that the company claims *cough* that it is capable of identifing and tracking of billions of existing and new MP3 files on the internet providing (get this) exact accounting for the copyright. "Thus enabling file sharing and linking value added data to songs" Fucked said in a pathetic attempt to spin. When asked if they were suggesting that it is currently illegal to share files, Fucked said "No Comment."
Fucked also announced that, in order to cover it's massive burn rate, it has duped some brainless Europeans (similar to brainless Americans, but know more than one lanugage) into throwing money at Fucked. Musican Eric Clapton has been starving in recent months due to the evils of Napster, but still managed to scrape up a few million dollars to throw into the furnace. "Mr. Clapton's investment in the company speaks of the importance importance Music DNA will have in returning to the record labels their rightful monopolies, I mean, I saw the guy, he's all skin and bones." Someone said in another interestingly unattributed quote.
The company anticipates that with industry-wide adoption of its music registry, acceptence by every node on the internet, a constutional amendment, a UN Resolution, and a few minor acts of God, the system will enable copyright holders to identify their content usage through at least a portion of the internet, thus ensuring that ownership and royalty right are fully "exploited, oops, don't print that, I meant 'monetized'". According to Fucked, Music DNA dosen't have an offical ship date but should come out "in a few months".
Music DNA is an extension of other FuckedFromTheOutset products which have already made a huge impact on the distribution of copyrighted material across the Internet, which include a bunch of neat sounding jargon and buzzwords. "I assure you, we have tons of buzzwords. MCSE's bow to our buzzword dominance".
FuckedFromTheOutset bullshits about how the process works: "Ok, see, it's sorta like this, Songs have patterns, right? and these don't change much if you have an exact digital copy, like a compressed 40kpbs mp3 recorded throught an analog bridge, see? So bitrate dosen't matter because this is about the information carred in it, all codec's have the same information, they don't try to elimanate information and guess at what's in the gaps." our weakly attributed source continued making a fool of himself for a few minutes, then said "Search engines can increase by atleast tenfold the amount of time and bandwith their spiders crawl through to make sure they're not linking to copyrighted materal, they're really gung ho about that, plus, an analyzer can be incorperated into a peice of client software residing on the PC to er, make sure the music is complete? Appearently, one can't figure that out by listening to it. We've talked with the XMMS people, they're all over that."
Mor E. Assplease, an investor in the company fumbles: "Obviously, copyright protection rackets maintainence is a seminal issue confronting the Cyber-eNew iEconomy.com at the moment, and music is at the heard of the matter. With Music DNA, Napster and Scour could cover their asses by putting a lame block that dosen't work to appease the courts. We can now account to the artists and songwriters who have been shortchanged by the labels for long before the eInternet iEconomy.com, or wait, I didn't mean that". The company's Olsen Wells expresses his hopes for the process, adding that "as the industry transitions from music as a product to music as a service, Music DNA could conceivably have the greatest single impact on the music buisness since the creation of the MP3 format". When asked if he could clarify that statment, remove a few buzzwords, or somehow make it make sense, Wells replied "No Comment".
Richard Stallman, leader of the Free Software Foundation, and proponent of free music, corrected our use of the word 'Linux' (appearently, it's GNU/Linux) but then began to laugh hysterically as we attempted to explain what Music DNA was. "I can just mess it up with dd on my Linux box" He continued, "GNU! GNU/Linux box I mean! please don't print that".
Lawrence Lessing, a Technolibertian known for his book Code and other Laws of Cyberspace, when asked about it, faught to keep an amused look off of his face and said "Well, we've obviously overestimated the enemy here, I'll have to drastically restructure my 'invisable hand' theory, it assumes a much higher caliber opponent than that with which we are dealing".
Jordan Bettis
``Wherever you go, there's another stupid sigfile quote.''An argument similar to this was used to get the mandatory-porn-filters-in-schools-n-libraries amendment included in the House Appropriations bill that has a good possibility of being passed in the next week or two:
- Mr. McCain: Internet filtering system work[s], and they need not be blunt instruments that unduly constrain the availability of legitimately instructional material. Today they are adaptable, capable of being fine-tuned to accommodate changes in websites as well as the evolving needs of individual schools and even individual lesson-plans.
...
From porn filtering to copyright filtering. Not a large leap.As we have seen through an increasing flurry of shocking media reports, the Internet has become the tool of choice for pedophiles who utilize the Internet to lure and seduce children into illegal and abusive sexual activity. ...As we wire America's children to the Internet, we are inviting these lowlifes to prey upon our children in every classroom and library in America.
--
or any other lossless compression mechanism. Wouldn't be too hard to develop 3pm, either, which stores the files backwards, or shuffles every n seconds, where n is a number between say 1 and 10 (depending on bandwidth).
When will they realise (like BMG) that working with this new paradigm is much better than trying to defeat it? Oops, I guess they still haven't figured it out, witness the losing 'War on Drugs'
"Those who would give up essential liberty for temporary safety deserve neither liberty nor safety" - Benjamin Franklin,
- A.P.
--
* CmdrTaco is an idiot.
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
Please. Anybody who thinks that's a word obviously has about 3 working brain cells (i.e. marketing.)
- A.P.
--
* CmdrTaco is an idiot.
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
If this works as it should, how is it going to distinguish between covers/bootlegs and the original? This is particularly important with bands like Pearl Jam allowing bootlegs to be distributed for non profit purposes freely on the internet.
If the file contains a brief obscenity, so much the better.
Those not running Apache, well, they need their own solution. Or to upgrade to Apache!
Warning: This signature may offend some viewers.
Their "application" (a webcrawler not logging 'illegal' mp3s) is a load of crap. Let's say I have cut in the first 15 seconds of a copywritten song--without permission--as a sample that I go on to critique in the audio file. I think that's fair use.
IANAL, but neither is the webcrawler a lawyer. It doesn't have the ability to judge fair use.
Worse, think if this 'webcrawler' is an RIAA bot looking for people to sue. It could lead to lots of frivolous actions.
Stevemp3, pron, are by far the two biggest search catagories out there, if the shlock-horror headlines in most rags is true.
"Dear Mrs search engine owner, please may we install something on your search engine servers to cut out a sizeable proportion of your customer base?"
What a great business model? it will -require- a law change to work, unless they think UCITA/DMCA can already be used to imtimidate big players like altaVista.
EZ
-'Press Ctrl + Alt + Delete to log on..'
"Oops, I always forget the purpose of competition is to divide people into winners and losers." - Hobbes
This could be a great system. A friend and me are looking for a way of synchronising our mp3 collection - but not copying the same song if theres a few seconds in difference in size, or a different name.
:)
this system could be very useful
Could be great for hours of fun.
I'm trying to teach myself to set people on fire with my mind... Is it hot in here?
It seems fairly easy to me do do a good "fingerprint" of a song by doing the math, determining the notes of the song, and the tempo, and maybe even determining who is singing based on voice sample matches once you're close.
#2. It's hard to defeat
Once you've got the code to do it, you can tweak the engine to work with different bit rates, streaming, etc.
Because they base it on the psychoacoustic model, it pays attention only to the parts you want to hear anyway. It will ignore the various means you use to tweak the files, as long as they sound the same, which is the main goal for the consumer of the files in the first place.
#3. It's hard/impossible to implement
What's also obvious is that the "search engine" would now have to download every instance of MP3 file it happens to encounter. This whould result in a massive increase in the amount of traffic for an already futile system of indexing the web.
We've already seen that the spiders that back search engines just don't have a prayer of keeping up with everything that is available. This is just dealing with the text part of web pages. Imagine trying to deal with millions of 3-10 Megabyte files that change every day!
#4. It must surface in a different model
It's just not feasible to download all of the MP3s that are available to do this, which means that the system is going to have to be selective in its downloading, and will, by necessity, result in "selective enforcement" of any laws this may detect the violation thereof.
If lawmakers decide to run with this approach, they'll have to settle for selective enforcement (with the resulting requirement of making the penalty huge to compensate for the odds of getting caught), or they will have to resort to the insipid approach of requiring ISPs to run the program against their own servers. (The FBI could also be even more insidious and build it into Carnivore). Let's also consider it might get built as a feature into the web servers. (Good thing Apache is open source!)
Mike Warot, Hoosier
How is this going to deal with gnutella, freenet, mojonation etc?
Me, I like the 'private networking' option in Gnotella and others. Me and my buddies setup private little sharing networks. I believe that Groove and others have taken this P2P thing to new heights also.
Sure, this may well work for all those geocities accounts and stuff, but at last count there were about, what? 20million+ Napster users ...
When will these turkeys wake up and stop trying to prosecute their customers.
This system is gonna have a really hard time with classical music. How can it know whether it's an amateur interpretation or a CD rip? The system can either accept everything or accept nothing, hereby insulting either the music labels or common sense.
Let's just say you have a server with a bunch of MP3's on it. And let's say this analysis of mp3's becomes a viable technology. Well then what is to prevent me from configuring *my* server (banning the ip) to ignore the search engine that implements this? :)
Steve
The only working method (taking things like freenet into account) that I can think of would be the closing up of both the hardware and software that's used in connecting to the net.
Integrate the network adapter into the motherboard and make it add a unique and traceable (who bought it, physical location, packet contents hint,...) ID into every packet.
No more self-assembled computers. Access to the stuff inside the chassis would be allowed only to authorized personnel. Just like heroin can be manufactured and sold legally today but only by the authorized people in the drug industry. Any unauthorized access would be a criminal offence.
Only authorized Operating Systems and device drivers allowed. Programming tools would also become controlled material.
Sounds to me like someone has found a creative way to make money off the fear that big record companies have towards mp3. Sell them some fancy system that will basically just be a big waste of time.
Let's think about just a few of the simple ways to defeat such a system.
Firstly - password protect mp3 download sites - Duh. In which case if the robot gets unauthorized access to the site, the ppl running it would be liable to break & enter charges.
Secondly, it would be a very simple matter to have an mp3 encoder shift a lot of the audio values around so that any track appears quite differently from the perspective of a binary analysis, but doesn't alter the end sound remarkably.
Yet another example of how AI isn't. And how it is always much simpler to fool an AI than it is to improve it. Think of the Iraqi techniques to fool american smart-bombs - current AI systems are all incredibly stupid when put against even moderate human ingenuity.
There are a thousand forms of subversion, but few can equal the convenience and immediacy of a cream pie -Noel Godin