March 2005Digital Fingerprinting for Digital Fingerprinting for Multimedia ForensicsMultimedia ForensicsÎ Include research sponsored in part by Air Force Research Lab and NSF; and contributions from Hong Zhao, Shan He, Hongmei Gou, and Zang Li.K.J. Ray Liu, Wade Trappe*, Z. Jane Wang**, and Min WuECE Dept., University of Maryland, College Park* WINLAB/ECE, Rutgers University** ECE, University of British Columbia, CanadaDigital Fingerprinting for Multimedia Forensics2Digital Fingerprinting and Tracing TraitorsDigital Fingerprinting and Tracing Traitorsz Leak of information as well as alteration and repackaging poses serious threats to government operations and commercial markets – e.g., pirated content or classified documentz Promising countermeasure:robustly embed digital fingerprints– Insert ID or “fingerprint” (often through conventional watermarking) to identify each user– Purpose: deter information leakage; digital rights management(DRM)– Challenge: imperceptibility, robustness, tracing capabilitystudioThe Lord ofthe RingAliceBobCarlw1w2w3SellSellDigital Fingerprinting for Multimedia Forensics4z Potential civilian use for digital rights management (DRM) Copyright industry – $500+ Billion business ~ 5% U.S. GDPz Alleged Movie Pirate Arrested (23 January 2004)– A real case of a successful deployment of 'traitor-tracing' mechanism in the digital realm– Use invisible fingerprints to protect screener copies of pre-release moviesCarmine Caridi Russell friends … Internetw1Last SamuraiHollywood studio traced pirated versionhttp://www.msnbc.msn.com/id/4037016/Case Study: Tracing Movie Screening CopiesCase Study: Tracing Movie Screening CopiesDigital Fingerprinting for Multimedia Forensics5Embedded Fingerprinting for MultimediaEmbedded Fingerprinting for MultimediaembedembedDigital FingerprintMultimedia Document 101101 …101101 …Customer’s ID: AliceDistribute to AliceFingerprinted CopyFingerprinted CopyembedembedDigital FingerprintMultimedia Document 101101 …101101 …Customer’s ID: AliceDistribute to AliceFingerprinted CopyFingerprinted CopyCollusion Attack Collusion Attack (to remove fingerprints)(to remove fingerprints)AliceAliceBobBobColluded CopyColluded CopyUnauthorized Unauthorized rere--distributiondistributionMFingerprinted docfor different usersCollusion Attack Collusion Attack (to remove fingerprints)(to remove fingerprints)AliceAliceBobBobColluded CopyColluded CopyUnauthorized Unauthorized rere--distributiondistributionMFingerprinted docfor different usersExtract Extract FingerprintsFingerprintsSuspicious Suspicious CopyCopy101110 …101110 …CodebookAlice, Bob, …Identify Identify TraitorsTraitorsExtract Extract FingerprintsFingerprintsSuspicious Suspicious CopyCopy101110 …101110 …CodebookAlice, Bob, …Identify Identify TraitorsTraitorsEmbedded Finger-printingMulti-user AttacksTraitor TracingDigital Fingerprinting for Multimedia Forensics6AntiAnti--Collusion FingerprintingCollusion Fingerprintingz Build anti-collusion fingerprinting to trace traitors and colludersz Potential impact: gather digital evidence and trace culprits– Deter unauthorized dissemination in the first place let the bad guys know their high risk of being caught– Support secure information sharing in government, military, and intelligence operations– Commercial use for digital rights management (DRM)z Traditional colluder tracing is designed for generic data– E.g Boneh-Shaw (1995)z Multimedia poses unique challenges and opportunities– Fingerprint should be embedded seamlessly and robustly– Multimedia-specific characteristics may be exploitedDigital Fingerprinting for Multimedia Forensics7Outline of the TutorialOutline of the Tutorialz Preliminaries – Review of Robust Embedding– Overview of Collusion strategiesz Orthogonal Fingerprinting – Collusion Resistance Analysisz Group-oriented Fingerprintingz Coded Fingerprintingz Data Embedding in Curves for Fingerprinting Mapsz Secure and Efficient Distribution of Fingerprinted VideoDigital Fingerprinting for Multimedia Forensics8Preliminary: Robust Embedding and CollusionPreliminary: Robust Embedding and CollusionDigital Fingerprinting for Multimedia Forensics9Background: Invisible WatermarkBackground: Invisible Watermark10011010 …© Copyright …Digital Fingerprinting for Multimedia Forensics10 Key Points:9 Human perceptual model for imperceptibility9 Use spreading and noise-like watermark for robustness and imperceptibilitySpread Spectrum Embedding ProcedureWatermarked ImageDCT Original ImageHuman Visual ModelWatermark wi~ N (0,1)IDCT αJNDxyiiiwJNDxy⋅⋅+=αDigital Fingerprinting for Multimedia Forensics11Detection of Spread Spectrum WatermarkDetection of Spread Spectrum WatermarkTNb=1b=-1z Detection can be formulated as a hypothesis testing problem.z Optimal detector can be calculated from assumptions on distortion (host media and noise from attacks).z If distortion is i.i.d. N(0, σd2) then optimal detector is a correlator:z Antipodal Code Modulation./(:blind-non;/:blind2222www)ywwyTdNdTNTTσσ−==TNpresenceabsenceH0: <(x + noise), w >H1: < y + noise, w > = < w + (x + noise), w >orH0: < (x + noise) - x, w > = < noise, w > H1: < y + noise - x, w > = < w + noise, w >Digital Fingerprinting for Multimedia Forensics12Fingerprinting upon SpreadFingerprinting upon Spread--Spectrum EmbeddingSpectrum Embeddingz Spread-spectrum embedding– Provide very good tradeoff on imperceptibility and robustness, esp. under non-blind detection– Typical watermark-to-noise (WNR) ratio: -20dB in blind detection, 0dB in non-blind detectionz Choice of modulation schemesOrthogonal modulation(Binary) coded modulationfor or jjuw=∑==Biiijjb1uw{}1,0∈ijb{}1±∈ijb# of fingerprints= # of ortho. bases# of fingerprints >> # of orthorgonal basesDigital Fingerprinting for Multimedia Forensics14Collusion Attacks by Multiple UsersCollusion Attacks by Multiple Usersz Collusion: A cost-effective attack against multimedia fingerprints– Users with same content but different fingerprints come together to produce a new copy with diminished or attenuated fingerprints– Fairness: Each colluder contributes equal share through averaging, interleaving, and nonlinear combiningColluded copyCollusion byaveragingOriginally fingerprintedcopies1/3Alice ChrisBobColluded copyOriginally fingerprintedcopiesAlice ChrisCut-and-paste attackDigital Fingerprinting for Multimedia Forensics15172 173 176174