Aaron Sigel Apple Security Team

Aaron Sigel

• Aaron Sigel

• Apple Security Team

Joey writes and records a song

• Joey writes and records a song

• Song distributed on some sort of media
• Joey (and music company) want to sell recordings
• But digital info is easy to copy, on most media

• What can Joey (and Music Inc.) try to do?

• Look for copies?
• Mark recording to make it easier to find copies?
• Restrict media so only certain devices can play it?

Legal landscape

• Legal landscape

• Copyright law, fair use, DMCA

• Examine or modify content

• Content hashing and copyright crawling
• Watermarking
• Fingerprinting

• Regulate use through special content players

• Apply complex policies, need tamper-proof platform
• Some examples
• MediaMax CD3: restrict access on software players
• DVDs: CSS encryption and hardware/software players
• Windows Media Rights Management
• Office Information Rights Management

U.S. Constitution (A1, S8, C8):

• U.S. Constitution (A1, S8, C8):

• "Congress shall have power . . . to promote the progress of science and useful arts, by securing for limited times to authors and inventors the exclusive right to their respective writings and discoveries”

• Intro of printing press in England in 1400s

• control (censor) publication of books
• maintain registry of legal books

• 1710 law to protect authors’ works

• prevent another person from re-producing a book and putting their name on it

Balance two competing objectives

• Balance two competing objectives

• Protect works so the author gets financial reward
• Promote access: progress of science, arts

• Gives exclusive rights for limited time

• Reproduce the work, derive new works, distribute copies, perform or display it publicly
• Extends to life of author plus 70 years

• Applies to

• “original works of authorship” fixed in tangible medium of expression
• literary, dramatic, artistic, musical, pictorial, architectural works – software? hyperlinks?

Legal use of copyrighted works for education, research, reporting, etc.

• Legal use of copyrighted works for education, research, reporting, etc.

• must provide transformative value

• Determined by four factors

• purpose and character of the use
• nature of the copyrighted work
• amount of the copyrighted work used
• effect on market value of copyrighted work

Anyone here get a letter?

• Anyone here get a letter?

• Music industry monitors file sharing
• Law specifies high minimum penalties
• Recipients usually offered a chance to settle for ~$3000
• See: http://www.eff.org/wp/how-not-get-sued-file-sharing

• Limitations of copyright law

• Coarse-grained protection, hard to enforce

• Next topic

• Technology used to help enforce copyright

Suppose we had a “content-aware” hash function:

• Suppose we had a “content-aware” hash function:

• H: {music}  {short strings}

• satisfying:

• 1. If M1 and M2 are two music clips (e.g. MP3 files) that play the “same” song then H(M1) = H(M2)
• 2. Given a clip M a pirate cannot create an “acceptable” clip M’ such that H(M)  H(M’)

• Is this realistic?

• Hash function must resist all signal processing tricks
• Do not know such hash functions exist
• some claim to have them

Web crawler looks for copyright violations

• Web crawler looks for copyright violations

• Use list of hashes of all copyrighted content
• Scans all web sites, file-sharing networks, etc.
• For every music file found, compute hash and compare
• If match is found, call the lawyers

• Problems:

• Hash functions unlikely to exist for music
• Does not protect against anonymous postings: publius
• Very high workload

DigiMarc MarcSpider

• DigiMarc MarcSpider

• Crawls web looking for pirated images
• May use watermarking? (next topic)

• MOSS (Measure Of Software Similarity)

• Detect plagiarism in programming assignments, web pages
• http://www.cs.berkeley.edu/~aiken/moss.html

• SCAM: N. Shivakumar, Stanford.

• Crawls web looking for academic plagiarism
• Several success stories:
• http://www-db.stanford.edu/~shiva/SCAM/scamInfo.html

Embed hidden watermark at the recording studio

• Embed hidden watermark at the recording studio

• Embed( M, I ): outputs a watermarked version of music M with the information I embedded in it
• Retrieve( M’ ): takes a watermarked music file M’ and outputs the embedded watermark I

• Watermark requirements (not necessarily achievable):

• Watermark must be inaudible (music) or invisible (video)
• Watermark should be robust: Given M1 = Embed(M,I),
• pirate cannot create an “acceptable” M2 with Retrieve(M2)  I
• To do this, watermark must resist all signal processing tricks - resampling, cropping, low-pass filtering, …

Second image has watermark inserted by DOS software “White Noise Storm”

• Second image has watermark inserted by DOS software “White Noise Storm”

Copyright crawler uses “Retrieve” algorithm

• Copyright crawler uses “Retrieve” algorithm

• Benefits:

• Copyright crawler does not need list of all copyrighted material
• No need for content aware hash
• Watermarking music “seems” to be an “easier” problem.

• But, some of the same problems as before

• Does not defend against anonymous postings
• High workload
• Need to mark with buyer or trace copy to culprit

Embed & Retrieve algs are usually kept secret

• Embed & Retrieve algs are usually kept secret

• “Security by obscurity” – not a successful approach

• Do robust watermarking systems exist?

• We don’t know the answer
• StirMark
• Generic tool for removing image watermarks
• Oblivious to watermarking scheme
• SDMI challenge:
• Broken: Felten, et al.

Basic idea:

• Basic idea:

• Embed a unique user ID into each sold copy
• If user posts copy to web or file-sharing network,
• embedded user ID identifies user

• Problem:

• Need ability to create distinct and indistinguishable versions of object
• Collusion: two users can compare their objects to find parts of the fingerprint

DigiMarc: embeds creator’s serial number.

• DigiMarc: embeds creator’s serial number.

• Add or subtract small random quantities from each pixel. Embedded signal kept secret.

• Signafy (NEC).

• Add small modifications to random frequencies of entire Fourier Spectrum.
• Embedded signal kept secret.

• Caronni: Embed geom. shapes in background

• SigNum Tech. (SureSign).

Aris Tech (MusicCode):

• Aris Tech (MusicCode):

• Rate: 100 bits/sec of music

• Solana (E-DNA)

• Used by LiquidAudio.

• Argent:

• Embed full text information.
• FrameBased: info. inserted at random areas of signal
• Secret key determines random areas.

Industry consortium (SDMI) considering four technologies for deployment in next-gen music and players.

• Industry consortium (SDMI) considering four technologies for deployment in next-gen music and players.

• We (Princeton, Rice, Xerox researchers) study technologies, find that they don’t work very well.

• We write a paper detailing our findings.

• Paper accepted for publication at conference.

Music industry claims that our paper is a “technology” whose primary purpose is copyright circumvention

• Music industry claims that our paper is a “technology” whose primary purpose is copyright circumvention

• Similar claim for oral presentation

• Threatens to sue authors of paper, conference organizers, and employers

• Seeks control over contents of paper

Music industry (RIAA, SDMI, Verance) threatens lawsuit if we publish

• Music industry (RIAA, SDMI, Verance) threatens lawsuit if we publish

• Conference organizers also threatened. We withdraw paper because of threats.

• We file lawsuit seeking right to publish

• After legal wrangling, paper is published

• We managed to publish, but:

• Months of effort by researchers lost
• Hundreds of lawyer-hours spent ($$$)
• Member of our team loses his job
• Eight-month delay in release of our results

Legal landscape

• Legal landscape

• Examine or modify content

• Content hashing and copyright crawling
• Watermarking
• Fingerprinting

• Regulate use through special content players

• Apply complex policies, need tamper-proof platform
• Some examples
• MediaMax CD3: restrict access for software players
• DVDs: CSS encryption and hardware/software players
• Windows Media Rights Management
• Office Information Rights Management

Traditional access control

• Traditional access control

• Owner of computer sets discretionary access controls

• DRM controls

• Owner of content sets usage rights
• Player owned by “untrusted user” must enforce usage rights
• Additional issue: copyright law allows fair use

Goal

• Goal

• Restrict use of music CD on computer

• Method

• CD contains autorun file that causes Windows to launch LaunchCD.exe, installs “SbcpHid” driver
• Driver prevents copying of restricted CDs

• Failures

• LaunchCD.exe will not run on Linux
• On Windows: hold shift key while loading CD

• Digital Millennium Copyright Act (DMCA)

• Forbids circumvention of copy protection mechanisms, and circumvention tools and technologies

CD contains copy protection software

• CD contains copy protection software

• Copy protection software protected by rootkit

• Rootkit detected by RootkitRevealer

Basic idea:

• Basic idea:

• Content distributor encrypts content before releasing it Release: C = EK[content]
• Decryption key embedded in all players.
• Player will only decrypt if policy is satisfied.

• Note: cannot prevent copying after decryption

• User can probe bus to sound card
• Unlike watermarking: watermark is embedded in content Propagates in cleartext copies of content

• Problem: what if one pirate uses reverse engineering to expose global key k ??

CSS: Content Scrambling System

• CSS: Content Scrambling System

• Used to protect DVD movies

• Each DVD player manufacturer i has key Ki

• Embed same key Ksony in all players from Sony.
• Every DVD movie M is encrypted as follows:
• enc-content = EK[M] where K is a random key
• EKsony[k] , EKphilips[K] , …
• About 400 manufacturer keys

DeCSS:

• DeCSS:

• Extracted key from Xing software player
• Could decrypt any DVD playable on the Xing player
• MPAA revoked Xing key: disabled all Xing players!

• Bigger problem:

• Encryption algorithm in CSS is based on LFSR’s
• Very fast: video rate decryption on weak DVD player
• Very weak: given one manuf. key, can get all keys

Embed a distinct key in every player

• Embed a distinct key in every player

• Every node v has an associated key Kv.

• Every player corresponds to leaf node.

• Key for player i: all keys on path from root to leaf i.

Initially

• Initially

• Encrypt all content with key at root
• Any player can decrypt content.

• When player i is revoked

• Encrypt content-key so only players other than i can decrypt.

When pirate publishes single key on Internet, MPAA knows which keys to revoke.

• When pirate publishes single key on Internet, MPAA knows which keys to revoke.

• What if pirate sells pirated players?

• How can MPAA tell which keys embedded in player?

• Solution: Tracing systems can interact with player and determine how to revoke that player.

• How? Take crypto class…

Distribute information in specific format

• Distribute information in specific format

• Player that knows this format controls action

• Control reading, playing, or copying content
• Guarantee payment in proportion to use
• Count number of times content is used
• Transfer payment to distributor

• No end run

• Must be impossible to use content without player
• Player must be tamper resistant

Attaching rights and making authorization decisions

• Attaching rights and making authorization decisions

• Enforcing decisions in a tamper-resistant software and hardware

Vocabulary

• Vocabulary

• Principals: Alice, Bob
• Resources: movie, picture, song
• Rights: play, edit, print
• Properties: manager, employee, trusted

• Licenses and grants

• license ::= (grant, principal)
• Principal p issues/says grant g
• grant ::= ∀x1…∀xn (cond → conc)
• If cond holds, then conc holds
• conc ::= Pr(p) | Perm(p, r, s)
• Pr(p) means principal p has property Pr
• Perm(p, r, s) means p is permitted to exercise right r over resource s

High-bandwidth Digital Content Protection

• High-bandwidth Digital Content Protection

• HDCP is a specification developed by Intel Corporation to protect digital entertainment content across the DVI/HDMI interface

http://www....com/crypt/drm/freeme/README

• http://www....com/crypt/drm/freeme/README

• The software distributed with this README file removes content protection from any Windows Media Audio file (.wma file) that uses DRM version 2 (as implemented in Windows Media Player version 7). …

• http://www...com/crypt/drm/freeme/Technical

• This document describes version 2 of the Microsoft Digital Rights Management (MS-DRM), as applied to audio (.wma files). The sources for this material are varied ...
• The basic components of MS-DRM involve use of elliptic curve cryptography (ECC) for public key cryptography, DES for a block cipher, RC4 for a stream cipher, and SHA-1 for a hash function. There is also a block cipher which I haven't seen before, used in the MS-DRM system to build a MAC, or keyed hash function.

Implementation Details: It is imperative to execute the following steps to neutralize the Freeme software breach… 1. Update the Content Header This procedure is performed by the organizations that package content. In this step the content packager will add an attribute … to the header of the protected Windows Media file. 2. Update the License Server(s) Each license issuer must update its license server configuration to ensure that:

• Implementation Details: It is imperative to execute the following steps to neutralize the Freeme software breach… 1. Update the Content Header This procedure is performed by the organizations that package content. In this step the content packager will add an attribute … to the header of the protected Windows Media file. 2. Update the License Server(s) Each license issuer must update its license server configuration to ensure that:

• It does not issue licenses to users who have the compromised security component on their PCs
• It can issue licenses to users who have updated the security component on their PCs.

• NOTE: if the license server is not updated (with the steps above) and an updated client (a client that has been updated with the new security component) makes a request to the license server, the license server will fail and generate an error to the client. 3. Trigger update of the new security component on the server side This step updates the license server so it can detect the version number of the DRM security component that is making the license request, and redirect it to an upgrade Web page if the security component version is less than "2.2.0.1".

Restricts playing, recording, sharing of files

• Restricts playing, recording, sharing of files

• Allows media to be shared among devices
• Allows others to listen to (but not copy) music
• Can burn audio CD, eliminates DRM protection

• How it works (overview)

• iTunes uses encrypted MP4 audio files
• Acquire decryption key by trying to play song
• player generates a unique ID, sends ID to iTunes server
• if not over authorization limit, server sends decryption key
• Decryption key is encrypted in iTunes to prevent transfer to another machine

Prior to buying content from iTunes Store

• Prior to buying content from iTunes Store

• User creates an account with Apple's servers and then authorizes a PC or Mac running iTunes
• iTunes creates a globally unique ID for device, sends to server, assigned to user's iTunes account
• Five different machines can be authorized.

When a user buys a song

• When a user buys a song

• A user key is created for the purchased file
• Encrypted using master key included in protected song file
• Master key encrypted with user key, held by iTunes and server

• Playing a song

• iTunes does not need to connect to server
• iTunes has keys for all tracks in its library

When a new computer is authorized

• When a new computer is authorized

• it generates a globally unique ID number
• Stores ID on Apple server (up to 5 devices)
• Server sends new machine entire set of user keys for all the tracks purchased under the account

Discovered attacks while building iTunes client for Linux:

• Discovered attacks while building iTunes client for Linux:

• QTFairUse grabs song data
• After unlocked and uncompressed by iTunes, dumps raw stream into container file,
• VLC media player, PlayFair, Hymn, JHymn
• intercept unlocked but not yet uncompressed song files, creating a small, ready to play, unencrypted AAC file.
• PyMusique, a Linux client for the iTunes Store
• requests songs from Apple servers and downloads them without locking them
• FairKeys simulates iTunes client
• requests a user keys from server, unlocks purchased songs