AACS, BD+, AND THE LIMITS OF DRM J. Alex Halderman Princeton → UMich.

Presentation on theme: "AACS, BD+, AND THE LIMITS OF DRM J. Alex Halderman Princeton → UMich."— Presentation transcript:

1 AACS, BD+, AND THE LIMITS OF DRM J. Alex Halderman Princeton → UMich.

2 DRM for HD Video Discs AACS BD+ AACS

3 A Case Study Why study AACS and BD+?  What is the state of the art in deployed DRM?  How do DRM systems fail in practice?  Do real-world constraints differ from our models?  Is research addressing the right problems? A work in progress…

4 In This Talk 1. AACS design 2. How AACS was initially broken 3. Why circumvention went commercial 4. BD+ design and hacks 5. Evaluation: Limits of DRM?

5 Content Scrambling System (CSS) DRM for traditional DVDs 1996  Toshiba and Matsushita  Secret design – security by obscurity  Severely gate-limited  Home-brew 40-bit cipher, based on LFSR  Each player-maker assigned one of 409 player keys  Each disc contains title key encrypted with each valid player key  Can revoke any player key if compromised

6 Attacks on CSS Algorithms reverse engineered from player apps  Motivation is can’t play DVDs under Linux  DeCSS: open source app decrypts video given disc key Cryptanalysis (Frank A. Stevenson, et al.)  Recover player keys w/ known plaintext 2 16  Recover disc key without plaintext 2 25 Complete failure  Every player key posted online within days  Attempts to legally suppress DeCSS unsuccessful  Can strip DRM from all DVDs forever with no additional hacking 1999

7 Advanced Access Control System (AACS) DRM for next-gen video discs (HD DVD and Blu-Ray) 2005 Major improvements:  Open design process, published specs, real cryptographers  Uses standard ciphers — AES  Fine-grained player key revocation employing NNL subset-difference method (can efficiently target individual devices)  Traitor tracing to identify compromised players

8 AACS Overview

9 Initial AACS Attacks  December 26, 2006 muslix64 posts decryption code to Doom9 forum  Implemented based on published specs  Doesn’t include any keys, but author claims to have found title/volume keys in memory of a software player  Motivation is can’t play movies on non-HDCP monitor

10 Initial AACS Attacks  December 26, 2006 muslix64 posts decryption code to Doom9 forum  January 12, 2007 first title key posted online, in the form of a riddle 2/Reavers are bad mmmmkay...Google 4TW! Mark Twain Intermediate School Restaurant & Lounge Cent Celtic Designs Dover Pictorial Science Online Special Feature Link Building Strategies Starlifter Solar periodicity Dawson's Creek Music Guide Decisions Duncan's F ways to market your small or solo business WBFF Olivia Quinn Food Stamp Leaver Dalmations CITI FM Skippyslist 239 -> EF 33 -> 21 50 -> 32 159 -> 9F 125 -> 7D 131 -> 83 141 -> 8D 154 -> 9A 112 -> 70 86 -> 56 136 -> 88 45 -> 2D 191 -> BF 102 -> 66 92 -> 5C 213 -> D5

11 Initial AACS Attacks  December 26, 2006 muslix64 posts decryption code to Doom9 forum  January 12, 2007 first title key posted online, in the form of a riddle  January-February 2007 volume keys extracted for all published discs  Improved attack method: attempt to use each 16-byte sequence from memory image as a key  Online key databases founded

12 Initial AACS Attacks  December 26, 2006 muslix64 posts decryption code to Doom9 forum  January 12, 2007 first title key posted online, in the form of a riddle  January-February 2007 volume keys extracted for all published discs  February 11, 2007 arnezami extracts a processing key from software player memory  Single key decrypts all current discs  Like title and volume keys, was not obfuscated

13 Initial AACS Attacks  December 26, 2006 muslix64 posts decryption code to Doom9 forum  January 12, 2007 first title key posted online, in the form of a riddle  January-February 2007 volume keys extracted for all published discs  February 11, 2007 arnezami extracts a processing key from software  February 24, 2007 ATARI Vampire extracts device keys from WinDVD  Public effort: < 2 months from scratch to device key

14 Industry Response and Limitations  February 24, 2007 ATARI Vampire extracts device keys from WinDVD  Keys leaked because software obfuscation failed to protect them – isn’t key revocation intended to recover from this?  Industry announced the compromised keys would be revoked on all future titles starting May 23 May 23 ???  Contracts with player vendors require 90-day notice for revocations  Possibly reasonable, but limits effectiveness

15 Industry’s Legal Response  Legal threats against sites showing processing key (09 f9 11 02 9d 74 e3 5b d8 41 56 c5 63 56 88 c0)  Online protests after Digg complies with takedown

16 Can You Own an Integer?

17 New Processing Keys  New processing key released same day as films that revoked the old key  90 day cycle seems to permit enough time to reverse engineer another player  Doesn’t have to be same player  Industry responds with proactive renewal, requiring all software to change keys every few months  Yet since then, at no time has any marketed disc been unplayable

18 Pirates of the Caribbean Antigua (trade dispute over Internet gambling)

19 SlySoft AnyDVD-HD Circumvention for the masses Only € 79

20 AnyDVD-HD Releases  Initial HD DVD support – February 2007  Blu-Ray support – March 2007  Response to revoked keys – May 2007  Partial BD+ support – November 2007  Full BD+ support – March 2008  Response to revoked keys – April 2008  Response to revoked keys – September 2008  All updates free to registered users  Uncertain which devices they’re reverse engineering Speculation: software players

21 A Market for Circumvention  CSS: Weak DRM completely broken Open source implementations widely available  AACS: Stronger DRM and working revocation Requires ongoing labor to attack  First to defeat latest refresh has period of exclusivity – profit  Repeated hacks = business model  Stronger DRM → Circumvention as a service!

22 1. Disc ID, Encrypted K V AnyDVD Serial No. Switch to an Oracle  AnyDVD-HD initially included processing keys  Competing circumvention products found it easier to reverse engineer AnyDVD than licensed players  SlySoft switched gears: ships with database of old volume keys plus server for new keys  Meta-DRM!  Advantage: traitor tracing more difficult (can restrict queries) AnyDVD-HD SlySoft Server Volume Key K V 3. (Antigua) 2. Decrypts with Device Keys (Secret)

23 Advanced Protection: BD+  Big idea: Players support a small VM that allows each disc to implement its own countermeasures  Based on the Self-Protecting Digital Content (SPDC) technology developed by Cryptography Research, Inc.  Acquired by Macrovision in 2007 for $45 million  Adopted by Blu-Ray but not HD DVD May have helped decide the format war  Unlike AACS, no public specification  Limited use so far, ~5% of Blu-Ray titles (mostly Fox)

24 BD+ Operation  Transformations – Alter content stream before decoding  Basic Countermeasures – Respond to attacks  Advanced Countermeasures – Execute native code Countermeasures

25 Public Attacks Against BD+  Significant progress this month on Doom9  Public effort to build a VM and debugging tools  Info. from R.E., patents, and other docs. (esp. patent application 20070033419 by Kocher et al.)  VM instruction set now known RISC, based on DLX by Hennessy and Patterson  Instruction Filter – 32-bit register XORed with each instruction prior to execution (protects against static analysis)  OS traps for symmetric and public key crypto, local storage, media I/O, native code execution, …

26 SlySoft Attacks BD+  November 2007: Partial BD+ support in AnyDVD  Convince software player a disc image is an original  March 2008: Full BD+ support in AnyDVD  Implements VM to support decryption  June 2008: Fox adds new countermeasures in release of film “Jumper”  7 days later, SlySoft updates AnyDVD to restore support Will future countermeasures buy more or less than a week?

27 Review: Strengths/Weaknesses  Standard crypto, designed by serious cryptographers  Fine-grained device key revocation  Multiple layers of protection, BD+ allows per-title countermeasures  Slow to revoke device keys: at least 90 days  PC software players ease reverse engineering  Renewal ensures market for circumvention

28 Prospects: Secret Weapons?  Sequence keys – improved traitor tracing  But… tracing probably not the weakest link  More aggressive BD+ deployment  But… so far seems to buy a limited window  Better protection for software players (TPM?)  But… limited by glacial deployment of Vista

29 Tracing/Revocation Overrated?  Traitor tracing  Small number of players are the usual suspects  Industry doesn’t even bother to use all its tracing abilities (instead does proactive renewal)  Key revocation  Slooooow / expensive in practice (seems necessarily so)  Attackers extracting new keys as fast as old ones are revoked  Not the weakest links today (obfuscation?)

30 What if… Faster Revocation?  Suppose perfect tracing and instant revocation  A new game:  When attacker compromises a device, can release selected title keys or publish device key  Studios decide whether to revoke the device key immediately or wait  Publishing the key: renders all past discs forever decipherable (can’t force attacker to retain state), but no effect on future discs after revocation  Rational studios may delay revocation anyway, to retain release window on some titles, since after revoked the attacker might as well publish it

31 Evaluation: Success or Failure?  AACS a clear failure  Virtually all discs copyable for past 21 months  Renewability seems to have no practical impact  BD+ … ?  Current breaks show won’t live up to hype (“secure for 10 years”)  Compare cost of building countermeasures to cost of breaking them  Industry has weapon of surprise: May help some films be secure for short release windows

32 Are We at the Limits of DRM?  AACS/BD+ designed by very smart people  Relatively few compromises  Yet quite limited success May be impossible to do much better within the model of disconnected players for content distributed on mass-produced physical media. Your predictions?

33 AACS, BD+, AND THE LIMITS OF DRM J. Alex Halderman Princeton → UMich.