HTML Site::Resources:: Copyright InfoCenter:: Defeating filesharing

Seeding and P2P terrorism

The growth of peer-to-peer (P2P) filesharing is nothing new. From the very early 1990s people were passing files around using the established routes of IRC, email and home webservers. The problem for the music industry is that the rapid growth in MP3-format digital audio coincided with the release of 'organized' P2P such as Kazaa, Morpheus, WinMX and so on. It's entirely expected (and acceptable) that if millions of people are able to create and copy digital music then they are going to want to share it too, or rather get it. Surveys suggest that about 85% of P2P users only ever worry about downloading files they want, rather than making files available for others. The Net is a global consumer group, not a global team of philanthropists. We, as music industry professionals, have to accept that the idea of 'free for all digital music' is not going to be stopped, and certainly isn't going to be derailed by a few hundred lawsuits. However, there is an interesting statistical fact about P2P that is often overlooked - it is very easy to sabotage as it relies on an incredibly small number of people to start the ball rolling.

There are over 100 million regular users of the Internet, looking at a little under 6 billion web pages and sending a billion emails a year. However, if you spark up a P2P client such as WinMX and look for a song from outside the top 50 then you tend to see between 5 and 100 hosts reported. The files that these have 'on offer' are usually based on only 5 or 10 original MP3 encodings and they just circulate about. As a result, a relatively small number of deliberate shares can dramatically influence the 'pool' of available files. This is due to a simple fact - the overwhelming majority of P2P users are either technically incapable of creating the original MP3 files, or can't be bothered. P2P relies on a lot of people holding copies of files that only a few people bother to create, and so it takes very little effort to become a 'major player' on P2P.

This is the concept of seeding. No record label wants to see 'good' MP3s of their catalog on P2P, as it directly reduces legacy sales - no matter what the advocates say, it's a proven fact. Put a 320K rip of a song on P2P and a subsequent single won't sell as many copies. The problem is that 'helpful fans' will often post these high-spec files instantly they are available - a major headache for labels releasing singles from a published album. By the time the second single is due for add, the album may have been out (and on P2P) for several months. You can reasonably assume in that time that anyone who wanted a copy has gotten one. If the label tries to prevent their track getting onto P2P then they'd be fighting a losing battle for two reasons - firstly of course it's impossible by nature to police a P2P network, as the files can reside on any of millions of home computers. Secondly, P2P is a very good marketing device and it's a cheap and effective way to get people listening to a new artist. The goal is to get the music everywhere but not damage sales - and that is where the idea of seeding came in.

Making a good seed

Seeding is the creation of an intentionally-disrupted version of a digital audio file. The file has the same name, duration, size, ID3 tags and dates as would be expected from an 'authentic' rip so people on P2P cannot tell what it is until they download it. The disruption usually involves insertion of extra noises, periods of silence or degradation of quality, and is done in such a way as to give the 'idea' of the track but not something people would prefer over buying the CD. It is possible to go overboard with seeding and remove the track entirely, but that defeats the marketing ability of the seed. You want most of the track to survive so that genuine fans can hear what the released version will be like, and maybe make them decide to buy it.

Seeding allows pre-release publicity, but is also an effective post-release control on P2P due to this issue of 'the few serving the many'. Part of our work involves creation of small networks of 'users' on P2P that share seeded files. If there are on average only 20 users on a network that show a copy of a track, then by creating 20 'seed users' you instantly enforce a 50% chance of any user requesting the track getting the seed. By carefully crafting the file specifications that are shared by the seed users you can imply that the other copies are probably seeded too, and put a lot of casual P2P users off their quest for a proper copy. It can take a lot of effort on the part of users to tell the difference between a good seed and an authentic rip, and by making the seeds more attractive (for example by sharing 196k and 320k files as well as 128k) you can distort the public image of the available pool of files very easily. Even when authentic files start emerging, the seed network can mirror them and make it appear a waste of time to download. The very nature of some of the P2P clients (such as WinMX and Kazaa) mean that the seeds self-replicate, as users downloading the files usually end up sharing them without knowing it. Once the fanbase has picked up the idea that this or that P2P network is 'full of bad copies' then they tend to stop looking.

Building teenagers

Setting up a seed network is not that hard, compared to the effort involved in altering perception of website availability, P2P is a walk in the park. At the most basic level a record label needs simply to plug a Windows PC into their public IP pool, run all the common P2P clients on it, load up the seeded MP3s and leave it. You have to make the seed machine attractive, so it must be on a fast network connection (at least a 2M SDSL). The majority of P2P clients don't reveal the IP addresses of users, so being on a traceable-to-the-label IP address isn't usually a concern. Here, we use a more active approach and host the files on groups of machines, some running multi-user versions of the P2P software. We change the usernames every few days and constantly scan the 'other files' being shared, changing the seed specifications to match what else is out there. If someone starts sharing a 'MyNewBand_FirstSingle_GENUINE_RIP.mp3' file then so do we. We also create groups of 'seed personalites' so each user shares other (seeded) files from bands that a typical fan would be expecting to like too. Our Jennifer Lopez seed personality wouldn't be expected to host hundreds of Manson and Death Metal tracks but a scattering of 50 Cent and Britney will make our drone look more like a teenager.

Web Seeding and safecasting

Website seeding is an entirely different issue, and vastly more difficult. The majority of labels avoid web seeding and instead use 'trailer cuts' - 30 or 60 second samples of tracks, given away free on official sites in advance of the add date. People know that's what they're getting, and it's what they expect. There is no easy way to run a seed system on a website, as by the nature of the beast a website is traceable and static - people go there once then avoid it once it's known to be fake.

Safecasting is the term for online delivery of media that is protected against easy unlawful use. We have to say 'easy' as it's technically impossible to prevent the copying and distribution of digital audio. Even with streams and licenced DRM at the very basic level a user can plug their speaker socket into the mincrophone socket of another PC and record the tune as they're playing it. Hackers take great delight in breaking the licence and encoding systems on 'protected' media and then making software available that allows conversion or copying of the files. This is of course all illegal under the DMCA, but the people writing the hacks are usually beyond the jurisdiction of the US Code. We have to accept therefore that anything posted to the Web will be copied. To some extent it's not a great problem - yes it's illegal, but it's also good advertising. What matters is making sure that the files cannot be used to create illegal offline media - counterfeit CDs and so on.

Broadcast radio has been safecasting for decades. Many stations use shorter 'radio edits' of tracks, so recordings from the radio aren't the same as from a CD. Of course quality has been an issue, but the growth of digital radio has removed a lot of that. However the most effective safecasting system is the good ol'fashioned excitable DJ. By talking over the intro and outro of a track, they make it impossible to record and pass off as an authentic file. This principle can be applied to online broadcast media just as easily. Deliberate voiceovers of tracks, such as an artist 'introducing' their new song, make a quantitative change to the audio without annoying fans. If a label wants to push out a new track but doesn't want to go with the old-hat trailer cut method, then one option is to overlay the track with an interview of the artists describing the lyrics, how much fun they had recording it, what they were smoking the day they thought up the bridge, and so on. It turns a potentially-dangerous full release of the track into a highly effective marketing tool with zero legacy media impact. Simple things are sometimes easy too.