Overpeer Spewing Bogus Files on P2P Networks

nimec writes "Zeropaid.com has posted news of a company called Overpeer which is the source of all the bogus mp3 files that are popping up on the various P2P networks. Zeropaid, in the news article, said: 'If you've encountered the "loop" files, in which a section of the chorus or hook is repeated over and over, you've been tricked by OVERPEER. OVERPEER are doing this with the full knowlege and consent of Interscope and Universal Music, in fact they are under contract to Universal and other major record labels, and will be doing a LOT MORE of this type of "interdiction" in the near future.' Right now this doesn't bother me because these bogus files are few, very spread out and it is easy spot them. I'm just afraid that over time people will keep downloading these bogus mp3s and become too lazy to delete them, like they are when it comes to incomplete songs."

Comments

This disgusts me.

[Warmth Is Life]

There's nothing more annoying than finding a brand new album in a high quality bitrate and then finding out it's nothing but a loop of two seconds. There's nothing more annoying than finding a brand new album in a high quality bitrate and then finding out it's nothing but a loop of two seconds. There's nothing more annoying than finding a brand new album in a high quality bitrate and then finding out it's nothing but a loop of two seconds.

Even though I'm not a big fan of copyright....

[forkboy]

That's the problem with running a service that's (for the most part) black market...when someone starts fucking it all up with counter-attacks, there's really not a lot of recourse.

I was thinking that a moderation system would work, if it's implemented correctly. For instance, once a person has been sharing X GB of files for, say, 2 weeks, they start getting moderation points....they can use these points to flag a file as being a dummy. (or just a shitty rip) If a user gets too many files modded down, he becomes unable to gain moderation points for a certain period. The sharing requirements will make it undesirable for RIAA droids to pollute the moderation system, since they'll have to be sharing material of their own. (and any dummy files they have will hopefully be moderated down...and if they ARE sharing valid material, well, cool, they're contributing to their own demise)

Please, nitpick at this suggestion, I'd like to see if it's feasible or not.

Re:Even though I'm not a big fan of copyright....

[gengee]

It's actually a bit of a complicated problem though. For instance:

How do you know how long someone has been online? What stops the client from simply reporting they've been online since January 1st, 1970? You can't really trust the peers to whom they're directly connected to know either, because in a P2P network people constantly drop on and off.

How do you stop Overpeer and like-minded companies from lying about the moderation points? Why can't they give it +100, CD Quality?

The only solution I have thought of is rather slow and clumsy. Basically everyone gets unlimited moderation points...instead of incrementing the count, you simply say "This file is good" or "This file is bad". When the file is downloaded, the P2P client creates a small hash of the file and stores that hash, along with the filename and moderation of the file. Then during the search process, you do 2 searches. First you search for a filename. Instead of all the clients returning "Yes, I have that file" they return "Yes, I have that file, with a hash of: 34232SFDSFSDSDSD2323DSD". Then a search is done for all the hashcodes returned by the first search asking for everyone's moderation on that hashcode.

Then you give that file a percentage-score (i.e., 95% of users say this file, with this hashcode is bad) or 92% of users say this file, with this hashcode is bad.

But the solution won't really work, because it exponentially increases the amount of bandwidth/cpu time required to do a file search.

Anyone else have any ideas?

Re:Even though I'm not a big fan of copyright....

[josh crawley]

---"That's the problem with running a service that's (for the most part) black market...when someone starts fucking it all up with counter-attacks, there's really not a lot of recourse."

Yeah there is. You fight back. No holes barred type of fighting too. If you can catch him in the act, do shit , like ping floods. It's effective in cutting bandwidth 1 way.

---"I was thinking that a moderation system would work, if it's implemented correctly."

---"For instance, once a person has been sharing X GB of files for, say, 2 weeks, they start getting moderation points....they can use these points to flag a file as being a dummy. (or just a shitty rip) If a user gets too many files modded down, he becomes unable to gain moderation points for a certain period."

Already incorrect implementation. I'd simply have a writable part of the P2P fs that allows you to GPG sign a file. You sign the MD5 sum to your 'nick'. If it's good, you sign. If bad, you dont. Now if some idiot is signing bad shit, you can assign trustworthiness to 0. You could also apply 'trusted' user signs to other known good MD5 sums (from untrusted users).

This system creates a "Web of Trust" that cannot be spoofed. No moderation point system will ever cut it (since it relies on a server-no reason to)

---"The sharing requirements will make it undesirable for RIAA droids to pollute the moderation system, since they'll have to be sharing material of their own. (and any dummy files they have will hopefully be moderated down...and if they ARE sharing valid material, well, cool, they're contributing to their own demise)"

First, even 1 screech is enough to 'kill' a file. For example, in Cool Edit plugins, they inset a bell after 30 seconds. Very effective. Also, might I remend you that it's legal for the RIAA to warez these files. Who's gonna pick on them?

Please, nitpick at this suggestion, I'd like to see if it's feasible or not.

Re:Even though I'm not a big fan of copyright....

[mlinksva]

Bitzi offers a solution similar to the one proposed in the parent's parent(? file ratings and other metadata associated with full file hashes). For partial/subrange verification, check out the proposed Tree Hash EXchange format.

I don't have any sympathy...

[kinko]

... for people who download these thinking they are downloading the "real deal". At least the studios are using technical means and not legal means to attack those who break copyright (no I won't use the "p" word).
People who download songs and movies continuously only make bandwidth more expensive and/or capped for the rest of us.

I think it's kind of funny - we waited overnight to download "TPM" only to discover it was "Pearl Harbor" with the title changed.

YAWS

[ImaLamer]

I've got yet another work around suggestion.

Your p2p application (which supports metadata, hashes etc) will wait to add a downloaded file to the "shared" section until after you view it.

This would cut down on some short divx'd files (which won't play "out of the box") bogus mp3 files (overpeer) and whatever else.

A system which flags files as "ok" could come under attack because overpeer could just flag their files "ok" as well.

The system I suggested above would only of course work with files downloaded, not files you have existing on your computer. Of course through the hash system you could be verified against other people.

Overpeer... create mp3's backwards from one-way hashes! Good luck you bastards!

Considering we already have hash systems in Gnutella apps... they can suck me.

Would a moderation system slow them down?

[Skapare]

What is needed to stop this is a moderating system which ranks the various traded products, as identified by their MD5 checksum signatures, according to some "measure of quality". By rank ordering, it cannot be used to entirely shutdown a trading network since everything would still be available. Products at 50 out of 100 would have received a ratio of good vs. bad moderations better than 50% of other products, and worse than the other 50% of products. It would not necessarily be a 50/50 good/bad moderation. Thus flooding of bad moderations across the board would have no effect, though it could be used to drive very specific classes of products down the list. But eventually, people would see the abuse and mod them back up. It would be sort of like moderation on slashdot, but everyone gets to play.

Now would it be possible to have selective moderation like slashdot has? Only a central authority could do that the way slashdot does. The big question would be judging who gets moderation points. As far as I know, on slashdot, it's almost entirely automated. With product trading, it would be harder to measure the quality by automation, so someone has to manually make the judgement calls and that brings some risks as well.

If individuals could be identified uniquely in some way, without the risk of exposing real identity, then meta moderation might work. One way to do that would be a slow rate of generating some kind of signed digital certificate that allows only so many to be generated at a time per network that receives it (and no personal identifying info included, and no records kept). Moderations and meta moderations would be signed by these anonymous certificates. You wouldn't know who moderated, but what you would know is that a group of moderations by the same certificate are probably from the same person and can be judged accordingly, good or bad. Excessive levels of moderation would also weaken your merit and derate your contributions.

Why this could be good...

[SmileyBen]

I'm surprised nobody has pondered the fact that this could be a Very Good Thing(TM). If they continue to do this, surely they'll be blowing big holes in any future court cases. They say "Napster [replace with future contentious system] can't feature songs which are copyright". Napster says "How do we tell?". Judge says "Fine, you have to filter by filename". Napster says "But wait a minute, half the stuff with filenames of copyright songs isn't those songs at all". The fact is, by engaging with these networks, even to undermine them, the record industry damages their own court defence. Basically they will single-handedly prove that these networks aren't just for exchanging copyright material which you might not have the right to do, but for just about anything. When a court realises that, their case is blown to hell... ...I guess it's wishful thinking to imagine they would notice, though...

Predators are good for an ecosystem

[Cryogenes]

Let the RIAA take out those services which are too weak to defend themselves, it will only make the others stronger.

It is possible to design a filesharing service that defends itself against bogus files.

It is possible to define a protocol that hides the file lists of individual users.

It is possible to build CDRs that play, copy and rip copy-preventing CDs.

The pressure exerted by RIAA will turn these possibilities into realities - simple Darwinian evolution.

Re:Predators are good for an ecosystem

[javilon]

Indeed,

And this is an interesting software engineering problem. It is the first internet protocol that has to be designed from the ground up for anonymity and resilience. And that will grow in a hostile enviroment.

The TCP/IP stack was designed for resiliency and they did a good job, but this has to be even better, and we don't have the goverment on our side!

There are a couple of attempts at this. One is www.freenetproject.org (that seems to be stalled) and the other one is gnunet.

GNUnet is a decentralized network with confidential and authenticated communication. A first service implemented on top of the networking layer allows anonymous distribution and retrieval of content. GNUnet supports accounting to provide contributing nodes with better service.

So, they are wasting my bandwith!

[twitter]

Hey, where are all the bandwith trolls when you need them?

You don't really think that this is going to work do you? People will simply be annoyed and have to share more. Someone is going to have to pay for the increased bandwith usage and it's not Universal Music. So, Universal is stealing from cable opperators. It's like spam, but they don't even hope to make money off it.

You have not even thought that people might be trying to share files that were intended to be shared and are NOT owned by Unviersal Music. But that's like the big 5 music publishers, "No one but us can record music, right? Drool, Drool."

twitter, who has never bothered to download silly mass produced comercial music, is annoyed that Universal Music is going to waste his time. Universal, you suck.

Re:So?

[whopis]

couldn't these DoS attacks be considered illegal

I think the problem with that argument is that this really isn't a DoS attack. They are using a P2P file sharing network to share files. That's the purpose of the network. Just because it is a file that you don't want doesn't mean that it is a DoS attack.

web of trust

[medcalf]

This method only works as long as all sites are equally trusted. If p2p software develops the idea of a web of trust, this method will fail quickly. Basically, a web of trust allows a user to mark a site as trusted or untrusted. You trust sites that sites you trust trust. In other words, I mark my client to trust foo.net and bar.com, because they always provide good stuff. They trust me as well, and a few other sites like fubar.cc. Since one or more of my trusted sites trusts fubar.cc, I trust fubar.cc.

Eventually this evolves such that sites which post bogus music, low-quality rips and the like will not get used, because no one will trust them. And a good web of trust allows you to see the trust path that led you to a server, so that if you get something bad you explicitly can mark as untrusted the nearest site to that (since they didn't do a good screening job) even though they would otherwise implicitly be trusted.

Media companies and technical counter-measures

[br00tus]

I am a Gnutella developer and contributor. I guess I'll split this comment into two parts - how I feel about this, followed by a technical explanation of how Gnutella and other p2p networks do and will handle this. P2P is attacked in many ways and this one does not bother me that much because it is only affecting material they hold the copyright to. Nonetheless, even though I perceive this as a minor problem, I do perceive it as a problem to be dealt with. I have an idealistic notion about p2p, that it will be used as a free, open publishing medium so that costs, in terms of bandwidth and so forth, are paid by the consumers, not by the publishers. I'm realistic enough to realize it is used primarily for trading Britney Spears mp3's, Warcraft III zip's, avi's of the Matrix and mpg's of Alley Baggett's Playboy videos. I don't mind this, but I am hoping it helps take publishing out of the hands of a few corporations, and I believe this is what the long-term planners of the corporations who fund the RIAA and MPAA really fear. My chagrin in aiding those sharing material copyrighted by corporations is more in aiding the spread of corporate published crap than in any respect of so-called copyright that these billion dollar multinational corporations hold. I hate large multinational corporations, their executives, and the people who own those corporations (the majority of stock and bonds are held by a tiny rich elite of heirs. I would like to diminish their power by any means necessary. I think the best way of doing this however is creating an alternative (p2p) to their publishing empires.

So as I said, I do see this as one of the problems to be solved, although I feel it's of lesser importance. There are many ways of doing this. One of them is previewing - when downloading an audio or video file, when you're about 100k into it (100-200k if it's video), do a preview and see what you're getting. With this looping stuff you have to go farther than 100k however - preview one fourth to one third of the way into the audio files. Many Gnutella clients have a preview feature, as does Fasttrack (Kazaa).

Another method is to ban IP's and IP ranges spreading this. This is already being done - it's only a minor fix because they will always get around it, but it will help somewhat, they won't be able to have big servers spewing this stuff 24/7

The real way to fix this however is hashes. Which are already ubiquitous - they already exist and are known on Gnutella (Shareaza, Gnucleus, Morpheus, Bearshare, Limewire), Fasttrack (Kazaa) and Edonkey2000. On Gnutella (Shareaza) and Edonkey2000, you can click through or cut and paste these URI's (URLs) to files from web sites (or Usenet, IRC, e-mail, instant messengers, whatever) and start searching and downloading the files - for FastTrack (Kazaa), it is a little bit more time-consuming and complex, but worth it if you're going to be downloading a large file. The hash technology is already there, the key now is finding a trusted source for hashes which are both good and whose data is findable and downloadable on p2p networks, and for those sources to survive. I guess I'll detail how this is currently working with the various p2p networks, why not?

There are four major p2p networks - Gnutella, Fasttrack, Edonkey and Freenet. Freenet is a publishing network, the others are all file sharing networks, which is what we're concerned with. Gnutella and Fasttrack are the two largest networks. Edonkey2000 specializes somewhat in large files however, so if it's 100MB+ files you're after, Edonkey2000 is on par, and perhaps better in some ways currently, than Gnutella and FastTrack. Edonkey2000 and FastTrack are closed networks - closed source server/clients and closed protocol networks. Gnutella is open, the protocol is open, and robust open source server/clients like Gnutizen exist for it. This gives Gnutella advantages, such as a choice of multiple clients for virtually every platform, as well as other advantages. Of all the file sharing p2p networks, Gnutella is my favorite and I believe Gnutella is the future of p2p. I think competition amongst p2p networks is healthy however as every can steal everyone elses best features and innovations.

Gnutella files are hashed for HUGE with an implementation called sha1. You can read about the technical aspects here if you wish to. These hashes are useful for finding additional sources for found files so that one can resume downloads or download from multiple sources with integrity. Actually there's one caveat to that - if you are downloading from an honest client, it will tell you a truthful hash of it's data. A client could give a fake hash and then send other data - but you would have to directly download from the rogue. How clients deal with this is even more complex - Gnucleus downloads overlapping chunks - it downloads 1-2000 from one source and 1950-3950 from another - if 1950-2000 do not match from both sources, it marks both chunks as possibly bad. You can read more details about this in Gnutella documentation and discussion groups.

Aside from this usage, these hashes can be used externally as well. Currently, Shareaza, which is a pretty good servent (server/client), is the only one from which URI's (URL's) can be cut, paste, and clicked through to from the web/IRC/e-mail etc. I'm sure clients like Gnucleus will have this ability in the future. If you had Shareaza installed, you could click on a link like this - which is an, I believe uncopyrighted, Chomsky speech, Shareaza would launch (if you don't have it already) and would ask you if you want to download the file or cancel. If you select download it would connect to GnutellaNet, search for the file, and if it found a host which has the file and which has upload slots open, would start downloading it. Actually, the Slashdot "allowed HTML" filters are pulling some necessary characters out of the above link, so you can't click through on /., although you can on a normal HTML web page. I can't post an URL that you can cut and paste either since /. forces a line break after 40 characters or so, if /. didn't do this and the below was in one line, you could have cut and paste it into Shareaza, I'll show it here for an example, imagine this was all on one line for you to cut and paste, or better was just a link to cut. You can do this on any HTML page, it's just the Slashdot HTML parsing messing it up -

gnutella://sha1:HXHSJ6ATN3LQCCIOBGUEWV5FFCKP2KBL/N oam%20Chomsky%20-%20Audio%20Book%20-%20Noam%20Chom sky%20-%20At%20Johns%20Hopkins%20University.mp3/

I would give the above link a rank of "7", because the last time I searched for it, 7 people replied they had it. I have several hashes with a score of 80-90, meaning you're more likely to find or download them, but the above is the only one I have that I have enough confidence in that the data is uncopyrighted.

So now you have one link to a hash - where can you find trusted sources which tell you what hashes are ubiquitous, making it more likely you will find and be able to download them, are rated in terms of quality by multiple sources and so forth? Well for Gnutella, one source is Bitzi. You can search for data there, see what is the most reported, what things are ranked, see comments, see bit rates, file sizes, artists, titles and so forth. It is very cool. Most interaction is from Bitzi into Shareaza (the only Gnutella client that does this currently), but from within Shareaza if you find a file you can type "find Bitzi ticket" and see if the hash has been reported on already. One thing which I'm sure will soon be remedied is that Bitzi does not have direct clickthrough to Shareaza, I have to copy hashes to my clipboard, edit them to Shareaza format and paste them into Shareaza. I'm sure soon Shareaza and Bitzi will agree on a standard and remove this step so I can just click through. And soon Gnutella clients other than Shareaza will have this ability as well. Bitzi's data base is open to the public, you can read their open data policy on their web site, anyone is free to use the data as long as Bitzi is credited. Bitzi.com is the only large, good source of Gnutella hashes I know of. Edonkey2000 has had hashes for a while, and has several good, large sources for hashes such as Filenexus.com and Sharereactor.com. Since Gnutella is a larger network and it just implemented this ability, I'm sure it will have even more and larger sources in addition to Bitzi. And since Bitzi's database is open to all, if Bitzi goes down someone else can open the database up again somewhere else. I'm sure in the future, even the trusted rating system will become distributed.

Gnutella uses the sha1 hash, Edonkey2000 uses another, and Kazaa uses another. Web sites exist that centralize the hashes for these. I'm sure soon web sites will exist that coalesces and translates all of this. Gordon Mohr, who runs Bitzi, wants to see a universal p2p tag, magnet, which is agnostic about which p2p backend it is using. Why not? We can have a tag that we (more or less) trust, and can retrieve the data from Gnutella, FastTrack, Edonkey2000 or Freenet. It's a great idea.

I am less interested in other p2p networks than Gnutella but I'll discuss their hash and meta-data web sites a little. The most interesting one is Edonkey2000, which as I said, has come to specialize in large (100MB+) files, and which I have to admit is a pretty good way to download large files with some guarantee of integrity. There are two major meta data sites for Edonkey - Filenexus and Sharereactor. There are other sites as well. If you're looking for large files, they do a pretty good job currently.

Fasttrack (Kazaa) uses hashing, but the Kazaa client is not that friendly to this kind of thing. So Fasttrack/Kazaa is more of a pain in this respect than any of the others. Nonetheless, you can download a program called Sig2dat that helps you copy and paste FastTrack's UUhashes. The you can go to web sites that give meta data, rankings and so forth to these hashes. Kazaa/FastTrack is unfriendly to all of this so it is much more of a pain - you have to install files that help you do this (sig2dat), you have to restart Kazaa for every file you want to download in this fashion and so forth. With Kazaa, all of this is a hassle, it's much easier to do in Gnutella (Shareaza), Edonkey2000 and Freenet.

And lastly there is Freenet. Freenet has been using hashes since the beginning. Freenet is a publishing network, not a file sharing network. That is nomenclature - file can be and are shared on Freenet - from html pages to gifs and jpgs, to mp3's, to avi's, although Freenet is the last place you want to look for large files, Freenet's bailiwick is small files. Even a 4 meg mp3 on Freenet is harder to find and slower to download than any of the other 3 networks. Small files are the domain of Freenet - HTML pages and images. The Freenet protocol is more rich than the other protocols in many ways, thus you have more than just audio and video files going over it, you have third-party applications utilizing it, thus you have things like Fproxy (A world-wide web equivalent which runs over Freenet) and Frost and Freenet message board (Usenet equivalents - both for text and binaries). One benefit of Freenet is it's hard to crack down on people for publishing information - because no one knows who data is coming from or going to. This is not absolute, but it is much safer than the file sharing p2p networks in this respect. Also, people publish data, so that what you put out is stored somewhere other than your computer, and if your web site or shared file or whatnot is popular, it will be out there all the time without your node needing to be connected. Freenet also used a lot of signatures, encryption and so forth, so you already have a pretty solid trust mechanism and data integrity. It depends on what hash is used - KSK hashes are insecure, but SSK are signed. So with Freenet there are large upsides and downsides - the downsides are downloading is much slower, since you're downloading via intermediaries, not directly, and the larger the file, the slower the download and the harder it is to find a complete file. The upshot of Freenet is that there is less of a legal risk with regards to sharing/publishing data, data is signed by the publisher which greatly helps integrity, and also Freenet's protocol allows extensions other than file sharing with it's own internal network - web and Usenet like applications, and I'm sure there will be more in the future.

Denial of Service, violation of Sprint AUP

[fmaxwell]

Overpeer.com is getting IP service through Telemerc who, in turn, gets service through Sprintlink.net. Accroding to the Sprintlink.net's Acceptable Use Police , the following are prohibited:

7. Knowingly engage in any activities that will cause a denial-of-service (e.g., synchronized number sequence attacks) to any Sprint customers or end-users whether on the Sprint network or on another provider's network.

and

9. Using Sprint's Services to interfere with the use of the Sprint network by other customers or authorized users.

That's practically a description of overpeer.com's business model. They use their bogus material to interfere with the use of P2P services and to effectively create a Denial of Service attack against P2P services.

I encourage Slashdot readers to contact Telemerc and Sprintlink at helpdesk@telemerc.net and abuse@sprintlink.net respectively and explain (in a civil manner) that you wish them to stop providing services to Overpeer because of the DoS business model.