New version page

IMPROVING COLLUSION RESISTANCE OF ERROR CORRECTING CODE BASED MULTIMEDIA FINGERPRINTING

This preview shows page 1 out of 4 pages.

View Full Document
View Full Document

End of preview. Want to read all 4 pages?

Upload your study docs or become a GradeBuddy member to access this document.

View Full Document
Unformatted text preview:

IMPROVING COLLUSION RESISTANCE OF ERROR CORRECTING CODE BASEDMULTIMEDIA FINGERPRINTINGShan He and Min WuElectrical & Computer Engineering Department, University of Maryland, College ParkABSTRACTDigital fingerprinting protects multimedia content from illegal re-distribution by uniquely marking copies of the content distributedto each user. Collusion is a powerful attack whereby several dif-ferently fingerprinted copies of the same content are combined to-gether to attenuate or remove the fingerprints. Focusing on theerror correction code (ECC) based fingerprinting, we explore inthis paper new avenues that can substantially improve its collusionresistance, and in the mean time retain its advantages in detec-tion complexity and fast distribution. Our analysis suggests a greatneed of jointly considering the coding, embedding, and detectionissues, and inspires the proposed technique of permuted subseg-ment embedding that is able to substantially improve the collusionresistance of ECC based fingerprinting.1. INTRODUCTIONTechnology advancement has mademultimedia content widely avail-able and easy to process. Thesebenefits also bring ease to unautho-rized users who can duplicate and manipulate multimedia content,and re-distribute it to a large audience. Digital fingerprinting isan emerging technology to protect multimedia content from unau-thorized dissemination, whereby each user’s copy is identified by aunique ID embedded in his/her copy and the ID can be extracted tohelp identify culprits when a suspicious copy is found. A powerful,cost-effective attack from a group of users is collusion, where theusers combine their copies of the same content to generate a newversion. If designed improperly, the fingerprints can be weakenedor removed by the collusion attacks.A growing number of techniques have been proposed in theliterature to provide collusion resistance in multimedia fingerprint-ing systems. Many of them fall in one of the two categories,namely, the uncoded fingerprinting and the coded fingerprinting.The orthogonal fingerprinting is a typical example of uncoded fin-gerprinting. It assigns each user a spread spectrum sequence asthe fingerprint and the sequences among users are mutually or-thogonal [1][2]. An early work on coded fingerprinting focusedon generic data and introduced a two-level construction in codedomain to resist up to c colluders with high probability [3]. Thisbinary code was later used to modulate a direct spread spectrumsequence to embed the fingerprints in multimedia signals [4]. A q-ary ECC code resisting c colluders, constructed as the c-traceabilitycode or c-TA code in short, was employed and extended to dealwith symbol erasures contributed by noise or cropping in multi-media signal domain [5]. A recent code based on combinatorialdesign was proposed in [6], where each code bit is embedded inan overlapped fashion by modulating a spreading sequence thatContact information: {shanhe, minwu}@eng.umd.educovers the entire multimedia signal. The overlap spreading con-fines the types of manipulation from colluders, and colluders canbe identified through the code bits shared by them.Multimedia data such as audio and video often consist of a se-quence of naturally divided frames. Among the fingerprinting con-structions reviewed above, the ECC-based fingerprinting providesan inherent support of the frame structure. Owing to a relativelysmall alphabet size q compared to the number of users Nuas wellas one symbol being put in one non-overlapping media segment(which can be one frame or a group of frames), the ECC-basedfingerprinting has the potential to generate and distribute finger-printed media in an efficient way. For example, for each frame,a total of q copies carrying q different symbol values can be gen-erated beforehand; a fingerprinted copy for any user can then bequickly obtained by assembling appropriate copies of the framestogether according to the fingerprint code assigned to him/her. Thesmall alphabet size also keeps the computational complexity offingerprint detection lower than the orthogonal fingerprinting ap-proach [7]. Despite all these attractive advantages, ECC-based fin-gerprinting has rather limited collusion resistance, which is aboutone magnitude lower than that of the orthogonal fingerprinting inthe settings examined in our recent work [7]. The small alphabetsize serves as a double-edge sword here as it substantially reducesthe degrees of freedom in constructing fingerprint signals.The focus of this paper is to explore avenues that can both re-tain the advantages provided by the ECC-based fingerprinting andimprove the collusion resistance. We have observed that the ex-isting ECC fingerprinting works put most of the attention on thecode layer and few work has considered the interaction betweencoding and embedding. In the mean time, joint consideration ofcoding and embedding has shown promising results recently in [6]for non-segment based fingerprinting. This motivates us to ex-amine the interplay between the ECC code layer and the embed-ding layer. As we shall see, by employing a strategic embeddingmechanism referred to as the permuted subsegment embedding forputting the ECC fingerprint code into host media, we can benefitfrom the joint consideration of coding and embedding for ECC-based fingerprinting and substantially improve its collusion resis-tance.2. ECC BASED FINGERPRINTING SYSTEMSA typical framework of ECC based multimedia fingerprinting [7]includes an ECC based code layer and a spread spectrum basedembedding layer. An ECC codeword is assigned to represent eachuser, and embedded into the multimedia document with one sym-bol per segment. After the distribution of the fingerprinted copies,users may collaborate and mount cost-effective collusion attacks.In this paper we focus on two types of collusions. One is theinterleaving collusion, whereby each colluder contributes a non-overlapped set of segments and these segments are assembled toform a colluded copy. The other type is the averaging collusion,whereby colluders average the corresponding components in theircopies to generate a colluded version. Additional distortion maybe added to the multimedia signal, which is typically modelled asan additive noise [6].The existing works on ECC fingerprinting have primarily tar-geted at code-level collusion, which is equivalent to segment-by-segment interleaving. We take the c-TA code [5] as an example. Ac-TA code satisfies the condition that any


Loading Unlocking...
Login

Join to view IMPROVING COLLUSION RESISTANCE OF ERROR CORRECTING CODE BASED MULTIMEDIA FINGERPRINTING and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view IMPROVING COLLUSION RESISTANCE OF ERROR CORRECTING CODE BASED MULTIMEDIA FINGERPRINTING and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?