IF: Blu Ray wins format war THEN: Sony laughs all the way to the bank?

[inefficient]

Code is data

Unless, the key is stored in a random location in memory each time....

And if they really wanted to get crazy, randomly and out of order embed individual bytes of the key in a much larger value and change the decryption function to re-assemble the key value as a function of the decryption process in a way that never places the actual value in memory.

That's 2 ways I can think of to defeat the current method of aquiring keys.

And you were going to hide your ondemand updatable decryption process where...??

Hiding your data with confusing code is just obfuscation and not great security. It will be cracked.

 

[deathkiller]

The only way to prevent the key/content extraction is secure execution when no other process can access the decoder memory, I don't think that it can be done in Windows XP or in a standard PC CPU/[SIZE=-1]Motherboard in general.

I think that windows vista can make it very hard to do, preventing most people to break the system, but the P2P piracy will continue with HD Movies.
[/SIZE]

 

[DemoCoder]

But in this exploit it's not the player key that is compromised, rather their handling of the content's keys. So, say they patch the software players to handle the authentication process differently (keys no longer at the memory addresses where they used to be), at some point the patched player will have to use the content key to authenticate (the already compromised) content. Now, wouldn’t this already known key act as a pretty big signpost for someone looking to compromise the patched player just the same as the previous one?

Players don't "authenticate" content, they decrypt it. When AACS revokes a key, future discs will be encrypted in such as a way that *every key in existence but the revoked one, can decrypt the content*.

For example, let's say there are 4 billion keys. Key #123456789 has detected to be involved in piracy. Now, SpiderMan-3 is released, and it is AACS encrypted such that all 399,999,999 keys can decrypt it, except for Key #123456789. That means the player with that key can no longer pirate any movies.

In order to continue pirating, they will need another AACS. The only way to get AACS keys is from AACS, so they means purchasing something that contains the key. That means, to pirate a movie is atleast as expensive as buying atleast one playback key.

The purpose of AACS isn't to stop someone from playing back already pirated content, it's to stop a crack from being a global crack for eternity, for every disc published, for the entire lifetime of the standard. That was the problem with CSS on DVD, once it was cracked, you could decrypt every new DVD published using the crack, and there was nothing the publishers could do about it.

AACS/BR/HDDVD/Publishers don't want someone distributing an enduser player that has been hacked for the average joe to rip content. They figure, when they find such programs floating around the net, they'll revoke the keys, forcing a recrack.

Hardware players could establish a trusted environment, and if AACS caught some fly-by-night Chinese firm making players that allowed piracy, they'd stop giving them keys. The issue is software players, and that issue will be ameliorated somewhat by the fact that obfuscation can prevent a global crack.

For example, if WinDVD stored its keys at offset #123 in some DLL, or always "near" some pattern of bytes, one could probably make an extracter that can pull the key from any executable. However, software players could easily move to a model wherein the secure component must be downloaded from the website, and the component is individualized for each person, tying the key to a credit card identity, as well as requiring the secure components to be manually disassembled each time and traced by expert crackers.

This would shift piracy to individuals who run fake credit card identities and have expert x86 cracking skills, and they would have to redo the same work everytime their key was tracked down as the culprit (another feature designed into broadcast encryption)

Coupled with very tough criminal penalties (like with Satellite piracy), I think it is a significant step above DVD. Will it prevent HD rips? No, no system will be 100% crack proof. The goal is to raise significant barriers to the average joe, and making it an activity that requires alot of time and expertise, which limits the population of people that law enforcement has to go after.

So sure, they'll be Russian and Chinese groups ripping away and putting HD torrents up, and in the end, either software players will either require trusted hardware chips (probably on the GPU card), or they'll just do away with third party pure software PC playback altogether, as the publishers know that the vast vast majority of people eventually consumer discs won't be watching on PCs, but on dedicated TV CE devices, which specialized HW. Media Center PC *HD* edition anyone? Requirements: Trusted Computing Module.

 

[Gubbi]

Players don't "authenticate" content, they decrypt it. When AACS revokes a key, future discs will be encrypted in such as a way that *every key in existence but the revoked one, can decrypt the content*.

For example, let's say there are 4 billion keys. Key #123456789 has detected to be involved in piracy. Now, SpiderMan-3 is released, and it is AACS encrypted such that all 399,999,999 keys can decrypt it, except for Key #123456789. That means the player with that key can no longer pirate any movies.

In order to continue pirating, they will need another AACS. The only way to get AACS keys is from AACS, so they means purchasing something that contains the key. That means, to pirate a movie is atleast as expensive as buying atleast one playback key.

But how do the content providers find out which key was used to decrypt the movie?

Cheers

 

[DemoCoder]

The only way to prevent the key/content extraction is secure execution when no other process can access the decoder memory, I don't think that it can be done in Windows XP or in a standard PC CPU/[SIZE=-1]Motherboard in general.

I think that windows vista can make it very hard to do, preventing most people to break the system, but the P2P piracy will continue with HD Movies.
[/SIZE]

One way to do it is to put it on the GPU card. A chip that does secure AACS decryption, and reencryption with a negotiated stream cipher could keep the raw codec data encrypted in video memory. The GPU would then decrypt on read and encrypt on write, sort of like DXTC compression, and it decompressed the data. Finally, the encrypted data is sent to the HDCP/TMDS for output, again, encrypted. Only the display monitor would be able to get access to the raw "plaintext"

This requires a GPU will slightly more general purpose DSP capability to deal with the CABAC/CAVLC parts of the algorithm (but maybe the g80 is up to it).

The only way to break this system, is to either break the display (read out decrypted info before it goes to LCD matrix) or break the AACS chip and get the keys. (the GPU would store no keys itself, as the stream cipher would use a negotiated key, and breaking a transient negotiated key would require a sophisicated man-in-the-middle attack)

I worked on tamper resistent co-processor chips at IBM a decade ago, and it is trivial to protect chips from drilling, xrays, ultrasound, and other techniques. Our coprocessors would self destruct wiping out their keys on detection of xray, drilling, cutting, and other attempts. And no, random cosmic rays would not set it off. (xrays are detected by the telltale heat signature caused by silicon absorbing them) Drilling produces unique vibrations which the vibration sensor was programmed to detect using a small database of recorded signatures. The chip was wrapped in a very very fine wire mesh coated in resin/epoxy among other things, and if you didn't set off x-ray detectors, vibration sensors, you'd most likely break this mesh when you started digging. The chip wasn't expensive to produce either, about $100 in '96, all custom stuff.

 

[Zaphod]

For example, let's say there are 4 billion keys. Key #123456789 has detected to be involved in piracy. Now, SpiderMan-3 is released, and it is AACS encrypted such that all 399,999,999 keys can decrypt it, except for Key #123456789. That means the player with that key can no longer pirate any movies.

Sure, but Key #123456789 is not compromised (if I understand this hack correctly). They find the decrypted key from the content and use that to trick the playback software to play the (decrypted) copy. The 'backup' program itself is just an implementation of the AACS decyption protocol implemented from spec (without the key decryption part that's not public). Currently, the pirated Serenity use the content key for the 'proper' serenity, so to prevent the players from playing the pirated one, they'll have to kill all legal copies of Serenity as well. I can't see that happening. By 'authenticate', I meant that at some point the software player will have to read out the encrypted content key, decrypt it, and use it to see if this is valid content that it's allowed to decrypt. Having a bunch of known data points to look for would eventually compromise the new player (and so on), obfuscation or not.

As for your other points: So, they need a secure environment to prevent people poking around in software player memory, but they don't (currently) have that. If they make it too cumbersome for the consumer, they'll just kill their own format.

 

[DemoCoder]

But how do the content providers find out which key was used to decrypt the movie?

Cheers

They take the movie, break it up into segments, and take each segment and embed a unique watermark, and encrypt it, N times with N different watermarks. This is called the "sequence key" technlque in AACS terminology. Thus, upon encountering segment 0, a player #1, may only be able to decrypt it using sequence key #3, for example, while player #5 may only be able to decrypt it using sequence key #9.

Each player will then have a unique "path" through the stream when decrypting, with player #1 say, decrypting segment #0 with key #3 and segment #1 with key #7 and segment #2 with key #2, while another player might have the sequence #9, #5, #1

Obviously, having BluRay's extra storage is nice, but they don't need to do this for the entire movie, just selectively for a few segments. Now, instead of needing to press a different disc for each player, they need to say, press 16 different masters, or maybe 256, which is a much more tractable problem.

In any case, the ripped movie will have a unique sequence of watermarks that either identify the device responsible, or atleast narrow it down considerably. After several rips on the same player device, it would be a certainty. (Note, you don't need to have multiple encryptions on a single disc, you can just partition the discs, with each cluster of discs with a different title key/sequence key)

The innovations in AACS are considerable. First of all, hierarchical broadcast encryption is a revolutionary breakthrough in cryptography, one that I would have said was impossible years ago, on the order of the invention of Public Key cryptography in power. It already has built in techniques for powerful traceability against blackbox illicit boxes. Secondly, the Sequence Key algorithm was invented, which works together with broadcast encryption, and watermarking to up the ante on anonymous attacks (rip-and-reencode) Usage of watermarking and multiple encryptions to aid forensics is not a requirement and left up to the discretion of the publisher AFAIK.

Modern watermarks survive recompression, filtering, and decimation, and are stegnographically secure (e.g. undetectable)

I'm telling you, the AACS spec is not your fathers copy protection. It was designed and vetted by the world's best cryptographers in the open, among industry and academia. They are not stupid nor blind to the threat models, and have considered pretty much all avenues of attack and designed the system to be resilient.

Anyone thinking that simply bypassing the player to get the plaintext of the video frames, and ripping and re-encoding that and sticking it on bittorrent will keep them safe from getting detected is in for a rude awakening. In the worse case, they'll just track down your device key and revoke it. In the more likely case, the device key will provide enough information for them to track YOU down. If the key is from a third party software player, trust me on this, there will be a paper trail linking you to it, if you bought it with anything but cash, that's assuming the software players don't require activation.

 

[DemoCoder]

Currently, the pirated Serenity use the content key for the 'proper' serenity, so to prevent the players from playing the pirated one, they'll have to kill all legal copies of Serenity as well. I can't see that happening.

AACS MKB is not designed to revoke keys for already published and compromised content. It is designed to revoke the ability to play future content. Thus, if Serenity is compromised, it will always be compromised. There is no difference between ripping the title key and ripping the plaintext content itself in this regard. Either way, if you distribute the title key, or distribute the ripped content, they'll be able to track you down if they use multiple title keys.

You're just parroting the BackupHDDVD "attack" which is not a break of the AACS system. AACS can only revoke a player after a title has already been compromised, and it can only revoke it's ability to decrypt future content. For already compromised content, the idea is that "the cat is out of the bag", there's nothing they can do about it, but they can however use it from an investigative standpoint to track down the player and person responsible for distributing that particular key or content online.

The whole "I'll distribute the content/title key!" attack was the central design point of the IBM "Renewable Traitor Tracing" paper which was where the AACS SKB system originated.

As for your other points: So, they need a secure environment to prevent people poking around in software player memory, but they don't (currently) have that, and if they make it too cumbersome for the consumer, they'll just kill their own format.

#1 they only need it for preemptive security for software players. AACS already has proactive future protection and retroactive tracing features

#2 it's only cumbersome if the vendors make it so.

Frankly, IMHO, AACS doesn't need it. It doesn't need revocation blacklists nor BD+ nor ROMmark either. From the standpoint of mathematics, the system is already robust against the threat models presented by p2p rippers.

 

[Zaphod]

So, three components: a) encrypted content, b) encrypted content keys, c) player keys; where b+c decrypts b to b´, which can be used to decrypt a). The compromised part being plaintext storage of b´ in player memory.

#1 they only need it for preemptive security for software players. AACS already has proactive future protection and retroactive tracing features.

OK. Edit: Thanks for the paper information. It was an interresting read.

#2 it's only cumbersome if the vendors make it so.

But, how?

Will they say 'no more HD-DVD on XP, too insecure'?; or 'All software players must be validated by CC'? I don't think they're willing to go there. Thus, the AACS scheme in itself may be as secure as hell (admittedly, parts of it are above my head), but exogenous business decisions will ensure that there will be insecure implementations out there.

 

[DemoCoder]

So, three components: a) encrypted content, b) encrypted content keys, c) player keys; where b+c decrypts b to b´, which can be used to decrypt a). The compromised part being plaintext storage of b´ in player memory.

No no, AACS is far more complicated than that. In AACS, the MKB (Media Key Block) is used by the player to calculate the Media Key (which cannot be calculated by the device if it has been revoked. Not just "hard" or subject to software hacking, but mathematically impossible). If a KCD is present, the Media Key is can't be computed without it. The KCD which is a per-disc bit of random data, usually stored in ROMmark. The media key is used to process the sequence key block, which is then used in combination with the volume id (per-disc or per cluster or per movie) to compute *another* key which is used to decrypt the title key.

Sure, the title key could be snooped from memory, or, the decoded movie frames could be captured from memory, however this just compromises a single movie. Using traitor tracing techniques, AACS-LA will compute the device keys of the player which was used to "snoop" the title key or rip the content. They will then blacklist those keys, or revoke the entire software player. Meanwhile, they will show up to arrest you. Thus, in order to do more snooping, you'll have to buy another software player to get another set of keys so as to steal another title key. This could get expensive.

I explained how they'll trace you in another message. The scheme has just enough confidence and low cost to publishers that I would not personally risk distributing ripped title keys or ripped content.

 

[Zaphod]

No no, AACS is far more complicated than that.

Sure, but for the discussion at hand that's basically how the current attack works. I read the paper you mentioned.

They will then blacklist those keys, or revoke the entire software player. Meanwhile, they will show up to arrest you. Thus, in order to do more snooping, you'll have to buy another software player to get another set of keys so as to steal another title key. This could get expensive.

I explained how they'll trace you in another message. The scheme has just enough confidence and low cost to publishers that I would not personally risk distributing ripped title keys or ripped content.

In theory. In reality, I expect there will be other vunerabilities in software players giving dedicated pirates access to several variations of the same player, not connected to themselves in any way. In the real world, dedicated software pirates won't use a player they payed for that's tied to their social security number, cc-number, or anything else that's more tracable than they currently are.

Reading your paper, it seems to me, that software players on unsecure systems are not viable at all as an cascade of revoked player keys will break the mechanism. Again, exogenous business decisions will ensure that there will be insecure implementations out there.

 

[-tkf-]

Will Hollywood react to this?

First pirated HD DVD movie hits BitTorrent
http://arstechnica.com/news.ars/post/20070115-8622.html

7 out so far.. hmmm

 

[-tkf-]

This is called the "sequence key" technlque in AACS terminology.

That doesn´t seem to be used in any of the current titles, or at least those that was "hacked".

http://www.freedom-to-tinker.com/?p=1110

Sequence keys don’t seem to matter as of yet. Discs are not required to use sequence keys, and indeed we have yet to see a disc that uses them.
...
The BackupHDDVD tool, as it is today, cannot cope with discs that use the Sequence Key mechanism — it uses only the per-disc volume keys and does not have or use any sequence keys.

 

[DemoCoder]

That doesn´t seem to be used in any of the current titles, or at least those that was "hacked".

http://www.freedom-to-tinker.com/?p=1110

Yes, it's optional, and surely likely to become used once the level of piracy soars.

 

[DemoCoder]

Sure, but for the discussion at hand that's basically how the current attack works. I read the paper you mentioned.
In theory. In reality, I expect there will be other vunerabilities in software players giving dedicated pirates access to several variations of the same player, not connected to themselves in any way.

You still don't seem to understand.

The "variations" on the player are irrelevent. Only the device keys are relevant. The device keys are generated only by AACS-LA via a secure mechanism. Unless pirates break into AACS-LA HQ and steel the root keys, they aren't going to be generating their own device keys.

It matters not that I can take WinDVD and turn it into 50 different versions. Software vulnerabilities are irrelevent. AACS broadcast encryption security does not rest on the security of the client.

In the real world, dedicated software pirates won't use a player they payed for that's tied to their social security number, cc-number, or anything else that's more tracable than they currently are.

And how do you propose they get the software with the keys? If AACS starts requiring all software players to be activated online or via phone, which they can do, there will be no way to obtain a device key without engaging in identity theft. How many pirates want to cross the line into financial fraud as well?

Then what, you'll start postulating money laundered offshore bank accounts run by Russian mobsters doing all the pirating.

They don't need to be perfect, they just need to be able to catch a few people and break some limbs. Mobsters don't need to catch everyone who cheats the mob, they just need to make it clear that you don't want to fuck with them and that there is a reasonable non-zero chance they'll find out.

 

[-tkf-]

Yes, it's optional, and surely likely to become used once the level of piracy soars.

If the keys are gathered from hardware players, like the Toshiba A1 that basicly runs a Linux then it would require a retrace of who bought the actual hardware wouldn´t it?

Anyway i could see the real bad guys in this, those that actual sell and make profit on stuff like this, buying a hardware DVD player like the Toshiba, or maybe 50 of them. And with a week apart pick keys and produce the last months new HD-DVD releases. Just brain farting here but...

 

[DemoCoder]

Yes it would, but I'm sure you're away that this information can be tracked. Store inventory systems can tell which person/credit card bought which unit, in fact, they SELL this info to direct marketing associations. And shippers/suppliers can tell which serial numbered units went to which stores. Trust me, I bet given the device keys, AACS can track down which model, which store, and then subpoena the store records to find out who bought that device. I doubt they would even need to subpoena records, most stores would hand over the record if faced with a supplier lockout (no more HDDVDs/players for you)

 

[-tkf-]

Yes it would, but I'm sure you're away that this information can be tracked. Store inventory systems can tell which person/credit card bought which unit, in fact, they SELL this info to direct marketing associations. And shippers/suppliers can tell which serial numbered units went to which stores. Trust me, I bet given the device keys, AACS can track down which model, which store, and then subpoena the store records to find out who bought that device. I doubt they would even need to subpoena records, most stores would hand over the record if faced with a supplier lockout (no more HDDVDs/players for you)

I think those that pirate stuff in this way would have no problem getting their hands on all the players they want, eveything from asking a random to buy one and earn a "10'er" , to buying them used (mental note, be carefull about buying anythin HD used, the key may have been used ).

The keys will be out there and as time goes by it will be even harder to keep track, the article i linked even had a suggestion on how to avoid this "feature".

 

[Zaphod]

Then what, you'll start postulating money laundered offshore bank accounts run by Russian mobsters doing all the pirating.

Oh, please. Let go of the conjunctive hyperbole.

If AACS starts requiring all software players to be activated online or via phone, which they can do, there will be no way to obtain a device key without engaging in identity theft. How many pirates want to cross the line into financial fraud as well?

See, that's the only postulation here. IF. Do you honestly think they will? Do you think that will fly with consumers? MS could get away with it for Windows, HD-DVD, currently, cannot.

I’m not arguing the technical merits of AACS here (nor in favor of piracy), just against your uninhibited faith that they’ll be able to implement all the aspects needed for it to be secure without hurting their business interests more than a bit of online copyright violation wil

 

[London Geezer]

And how do you propose they get the software with the keys? If AACS starts requiring all software players to be activated online or via phone, which they can do, there will be no way to obtain a device key without engaging in identity theft. How many pirates want to cross the line into financial fraud as well?

Then what, you'll start postulating money laundered offshore bank accounts run by Russian mobsters doing all the pirating.

That will be the day digital distribution will be the best way to obtain material.