Slashdot Mirror

This is what Audible Magic does. Exactly.

http://audiblemagic.com/index.asp

So google is doing it again?

The problem is that even after you do all this, do you actually make more money?

When the [MP|RI]AA make the Network Appliances being Licensed

How exactly does CopySense work ? I just read a PDF of their propaganda. Apparently an artist wishing to "protect his copyrighted works" registers in their database. Then CopySense makes a "media fingerprint" of the files - what do they mean by that ?

If they mean an actual audio fingerprint like MusicBrainz does, wouldn't the entire file need to be downloaded first ? So they're scanning the entire P2P traffic for relevant packets and recomposing the file on their own systems (including compressed files, since they could be an entire discography, hey, who knows ?), and *then* comparing the fingerprint ? Sounds really resource-expensive to me. Then on with bigger, more complex files (movies, HD movies...). Then they also say they can filter out porn/kiddie porn P2P traffic. Have they got fingerprints of THAT ? All existing porn movies ? Sounds like the largest porn collection on Earth. And you would only need to distribute the files with the video turned upside down and the audio playing backwards to defeat the system. Or are they relying on torrent file names for that ?!

If they mean an SHA1/MD5/whatever fingerprint, it's even less feasible, as they would need one for every possible encoding (MP3 CBR 128kbps, MP3 CBR 192 kbps, MP3 VBR... Vorbis... FLAC...) and compression (.gz, .zip, .rar, .7z, .bz2...) and combination of both.

Does anyone here know exactly how CopySense work, and IF it works at all ?

Please correct me if I'm wrong, but on Audible Magic's "CopySense Appliance" website (what I'm thinking this is all about), it lists how it can also block child pornography -- kind of setting their sites a little high there, aren't they? Would it just be searching for some kind of file name, or what?

Palisade Systems is a privately held company. They could have a large investment by the RIAA, but probably not.

Palisade Systems flagship product, PacketSure, is the core appliance base for Audible Magic's http://www.audiblemagic.com/ P2P appliances. Audible Magic is responsible for CopySense, which is targeted directly at finding copyrighted material traveling over the ether. CopySense is probably available from Palisade, but isn't a feature Palisade pushes. Palisade is currently pushing PacketSure into things like on the wire document analysis for text data. Audible Magic is pushing their P2P appliance into the RIAA realm of copyrighted works

Hi!

I work for "the man" http://www.audiblemagic.com/

Now with Blocking and Content analysis.

!Hi

I'd like to know how detection of copyrighted materials works, myself. Apparently these people: http://www.audiblemagic.com/ think they have something that works. The college I work for is going to do a demo of the product to see if it can live up to its claims. I certainly hope it can't, as I hate to see freedoms and liberties going the way of the dodo...

Audible Magic is another "tool' that performs TCP resets when it detects what it believes to be illegal file sharing.

This type of activity is introducing a frightening level of interference in the network as a whole. Imagine trying to troubleshoot network problems caused by tools such as these "accidentally" dropping network connections.

Interesting question. Audible Magic's IEEE article has some background info on your concern.

From a practical perspective, in the US, if you record and produce a sound-alike cover of a song, odds are that you won't have the right to distribute it freely. The original lyricist and composer are entitled to royalties. There may be a rare exception (if, say, Godsmack covered a public domain song from the 1850's and your band covered their recording with accuracty that managed to fool Audible Magic's algorithm) but this typically won't be the case.

Those are all excellent questions, but luckily the answers are easily at hand. I went to the Audible Magic web site and with just a few clicks found a datasheet and even an IEEE article which should at least partially address your concerns. Hope this helps.

Those are all excellent questions, but luckily the answers are easily at hand. I went to the Audible Magic web site and with just a few clicks found a datasheet and even an IEEE article which should at least partially address your concerns. Hope this helps.

Those are all excellent questions, but luckily the answers are easily at hand. I went to the Audible Magic web site and with just a few clicks found a datasheet and even an IEEE article which should at least partially address your concerns. Hope this helps.

Their site primarily promotes it as a network appliance, that "passively listens to all traffic on the network" and "block all P2P traffic or specify that P2P transactions are limited to a specified bandwidth" or "You may also choose to block only offending copyrighted works from being traded on your network". So it doesn't have to be installed on the client or server side. Any network provider, business, school, etc., could install this appliance.

I make these things. Hashing is for data. CopySense by Audible Magic. They listen to network traffic, identify audio file types, decode them to audio, and compare them with a database of audio fingerprints.

http://www.audiblemagic.com/copysense_appliance.ht ml

They don't use hashes -- they use fingerprints, which allow for much fuzzier matches. You might also note that they don't check encoded files, they check the audio stream. And though it reduces the capacity, it is possible to scan zipped files.

It can't scan a fully-loaded 1000 Mbit connection, but what is your actual Internet connection speed? 2 Mbits? 10 Mbits? Do you really think that it's unreasonable to scan this much data in real time?

Then again, it runs a FreeBSD, so linux users will probably claim that it doesn't exist.

CopySense by Audible Magic. I make these things. They listen to network traffic, identify audio file types, decode them to audio, and compare them with a database of audio fingerprints.

http://www.audiblemagic.com/copysense_appliance.ht ml

They don't use hashes -- they use fingerprints, which allow for much fuzzier matches. You might also note that they don't check encoded files, they check the audio stream. And though it reduces the capacity, it is possible to scan zipped files.

It can't scan a fully-loaded 1000 Mbit connection, but what is your actual Internet connection speed? 2 Mbits? 10 Mbits? Do you really think that it's unreasonable to scan this much data in real time?

Then again, it runs a FreeBSD, so linux users will probably claim that it doesn't exist.

This depends on what the fingerprinting technology is. Audible Magic's technology, for example, can supposedly generate fingerprints for audio files that are independent of specific encoding. I don't know whether they'd end up with a single fingerprint for all versions of a track, or a couple, but in either case the number of fingerprints would be much smaller than the number of hash values.

Also, the system doesn't need to block every single copyrighted song to be effective, because not all songs are equally popular. There are perhaps 1m copyrighted tracks,

There are perhaps 1m copyrighted tracks. That's probably still too many to block. Luckily, tracks are not all equally popular, so if you block the most popular tracks, that would be disproportionately effective. Perhaps the Rolling Stone 500 Greatest Albums of All Time could be a start. That's only, say, 10,000 tracks.

How hard could it be to transmit 10,000 numeric fingerprints, and block sharing of any files that match them? P2P networks are great at transmitting data, so it'd be easy to update the list. :-)

Other than a convenient way to gather multiple tracks together, to prevent these guys from doing this kind of stuff.

Yep, little old Winzip is the Sharpie for this expensive DRM.

As far as the fingerprinting goes, there is a project that has going on for about a year now (that I know) to start fingerprinting music that gets played on radio stations. I used to work at the radio station at Texas Tech University (ktxt.net, Your Only Alternative!) and we were one of the first twenty five stations in the nation to get hooked up with a system to do our charting automatically (to CMJ) by this system (from Audible Magic - www.audiblemagic.com) that listened to what was basically an audio stream and identified the songs. During the testing period they would call when they heard a song that didn't get identified. The calls tapered off pretty quickly, and we play some fairly obscure stuff, so I'm guessing that the fingerprint database must be getting rather large. I'm glad that they are making the fingerprint database available to the average joe!

Now I can start ripping all those compilation tapes my sister made.

"Anti-Piracy" database/recognizers are being marketed by more than one vendor. Here's Replicheck. I understand this technology will be applied to P2P networks too.