J. A. O'Sullivan, P. Moulin, and J. M. Ettinger. Information-theoretic analysis of Steganography. Proceedings ISIT 98, 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....attention has been devoted to WM from the information theoretic perspective. In the few information theoretic papers that do exist on this topic, capacity and error exponents of WM systems were characterized by viewing them as coded communication systems with side information [4] 5] 8] 9] [10], 11] The model in these works is essentially as follows: A secret message, encoded by a rate R block code of length n, is hidden in a memoryless covertext message within small degradation of quality, symbolized by distortion level D 1 . An active attacker may introduce additional distortion D 2 ....

....open. While the assumption that the attack channel is memoryless may appear somewhat arbitrary, we adopt it here primarily for reasons of simplicity. It should also be noted that memoryless attack channels are worst case channels in certain communication situations in the contexts of WM [8] 9] [10], and the jamming game [13] 14] In addition, as it will turn out, at least for the case of covertext available at both sides, memoryless channels are indeed worst case channels from the viewpoint of identi cation capacity. 2 Side Information at Transmitter and Receiver A De nitions ....

[Article contains additional citation context not shown here]

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, \Information theoretic analysis of steganography, " Proc. ISIT `98, p. 297, 1998.

.... ways to hide data and to remove hidden data [1, 8, 2, 3, 4, 5, 6, 7] However, the information theoretic analysis that describes the fundamental limits of any watermarking or data hiding system and can ultimately guide the design of efficient watermarking algorithms has been emerging only recently [9, 10, 11, 12]. Information theory yields results that are substantially different from those obtained so far in the data hiding literature. A typical approach in this literature is to assume that the data embedding and decoding functions take a particular form, and that a particular type of attack is used ....

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information-Theoretic Analysis of Steganography," Proc. IEEE Int. Syrup. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998.

.... authentication problems is just emerging, there has been a growing literature in the information theory community on digital watermarking and data hiding problems, and more generally problems of coding with side information, much of which builds on the foundation of [18] 19] and [20] see, e.g. [21] [22] 23] 24] 25] 26] 27] 28] 29] 30] 31] 32] 33] 34] 35] 36] 37] and the references therein. Collectively, this work provides a useful context within which to examine the topic of authentication. In this paper, we propose one possible model for the authentication problem, and ....

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information-theoretic analysis of steganog- raphy," in Proc. Int. Symp. Inform. Theory, Cambridge, MA, Aug. 1998, p. 297.

....the notion of random coding not only serves as tool in a proof technique, but is an inherent part of the model. Random coding is enabled by a secret key that speci es the particular codebook that has been drawn, which is shared by the encoder and decoder. O Sullivan, Moulin and Ettinger [14] and Moulin and O Sullivan [12] characterized the watermarking codes from the information theoretic point of view as a capacity game between the information hider and the attacker. They considered two models of watermarking systems: The private game, where the covertext is available to both the ....

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, \Information theoretic analysis of steganography," Proc. ISIT '98 , p. 297, 1998.

....Watermarking has attracted interest in recent years due to the ease by which data can now be reproduced and transmitted around the world, for example see [3, 4, 5, 6] and references therein. The information theoretic model of the watermarking game was introduced by O Sullivan, Moulin and Ettinger [7]. They formulated private watermarking as a max min game over conditional mutual information, and extended their approach in [8, 9] For similar models (but with somewhat di#erent distortion constraints) error exponents were studied in [10, 11] and identification capacities in [12] Information ....

....level, and a.s. stands for almost surely . We will also consider an average distortion constraint on the encoder, i.e. E [d 1 (U , X) # D 1 , 10) where the expectation is with respect to the covertext, the message, and the secret key; a similar average distortion constraint was considered in [7, 9]. Still other types of constraints have been considered, e.g. E [d 1 (U , X) U = u] # D 1 for all u [10] and Pr d 1 (U , X) D 1 U = u # exp( #n) for some # and all u [11] The latter constraint reduces to (9) for # = #. Unless otherwise stated, we shall focus on the a.s. distortion ....

[Article contains additional citation context not shown here]

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information theoretic analysis of steganography," in Proc. of ISIT, (Cambridge, MA), p. 297, 1998.

....distortions. As with a scalar Gaussian covertext [1] the capacity does not depend on knowledge of the covertext at the decoder. Unlike the scalar version, an attacker based on the rate distortion solution (i.e. optimal compression) is suboptimal. I. Introduction The watermarking game [1, 2] can model a situation where an original source sequence ( covertext ) needs to be copyright protected before it is distributed to the public. The copyright ( message ) needs to be embedded in the distributed version ( stegotext ) so that no attacker with access to the stegotext will be able ....

J. O'Sullivan, P. Moulin, and J. Ettinger, "Information theoretic analysis of steganography," in Proc. of ISIT, 1998.

....distortions. Both the public version of the game (covertext known to neither attacker nor decoder) and the private version of the game (covertext unknown to attacker but known to decoder) are treated. Surprisingly, the two versions yield identical values. I. Introduction The watermarking game [1, 2] can model a situation where an original source sequence ( covertext ) needs to be copyright protected before it is distributed to the public. The copyright ( message ) needs to be embedded in the distributed version ( stegotext ) so that no attacker with access to the stegotext will be able ....

....that the optimal A is a root of a cubic equation and hence a closed form solution for the capacity exists. Di#erent capacity results for yet another version of this game with expected distortion constraints and a decoder that knows the attack strategy (ML decoder) have been recently reported in [1]. ....

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information theoretic analysis of steganography," In Proc. of ISIT,

....we refer to as dither modulation. We evaluate both specific realizations of uncoded and coded QIM, and the asymptotic performance limits of coded QIM using information theoretic analysis. Other emerging information theoretic results on the digital watermarking problem are developed in, e.g. [14] [20] The specific organization of the paper is as follows. In Section II, we develop two useful equivalent models for the information embedding problem. In Section III, we classify traditional approaches to this problem, and in the process identify some of their shortcomings. Section IV ....

....forms an estimate of the embedded information based on the channel output . We focus primarily on the host blind case of interest in most applications, where is not available to the decoder, in contrast to the known host case, where the decoder can separately observe . See, e.g. [14] [17] for information theoretic treatments of some aspects of the known host case. Our interest is in decoders that produce reliable estimates whenever the channel is not too severe, where reliable means either that deterministically or that for sufficiently small .In 1 The vector is any ....

[Article contains additional citation context not shown here]

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information theoretic analysis of steganography," in Proc.

....we refer to as dither modulation. We evaluate both specific realizations of uncoded and coded QIM, and the asymptotic performance limits of coded QIM using information theoretic analysis. Other emerging information theoretic results on the digital watermarking problem are developed in, e.g. [14, 15, 16, 17, 18, 19, 20]. The specific organization of the paper is as follows. In Sec. 2 we develop two useful equivalent models for the information embedding problem. In Sec. 3, we classify traditional approaches to this problem, and in the process identify some of their shortcomings. Sec. 4 introduces the QIM class ....

....forms an estimate m of the embedded information m based on the channel output y. We focus primarily on the host blind case of interest in most applications, where x is not available to the decoder, in contrast to the known host case, where the decoder can separately observe x. See, e.g. [14] [17] for information theoretic treatments of some aspects of the known host case. Our interest is in decoders that produce reliable estimates whenever the channel is not too severe, where reliable means either that m = m deterministically or that Pr[ m #= m] # for su#ciently small #. In ....

[Article contains additional citation context not shown here]

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information theoretic analysis of steganography, " in Proc. of the 1998 IEEE International Symposium on Information Theory, (Cambridge, MA), p. 297, Aug. 1998.

....to this problem from the information theoretic perspective. In the few information theoretic papers that do exist in this context, attempts were made to characterize capacity and or error exponents of watermarking systems by viewing them as coded communication systems with side information [6] 7] [14], 13] 12] 15] In general, the system model that has been studied is essentially as follows: A secret message, encoded with a rate R block code of length n, is hidden in a memoryless covertext message (an image) within small degradation of quality, symbolized by distortion level D 1 with ....

....While the assumption that the attack channel is memoryless might appear somewhat limited, we adopt it here for reasons of simplicity. It should also be noted that memoryless attack channels are worst case channels in certain communication situations in the contexts of watermarking [12] 13] [14], and the jamming game [17] 18] In addition, as it will turn out, at least for the case of covertext available at both sides, memoryless channels are indeed worst case channels from the viewpoint of identi cation capacity. 2 Side Information at Transmitter and Receiver A De nitions Throughout ....

[Article contains additional citation context not shown here]

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, \Information theoretic analysis of steganography, " Proc. ISIT `98, p. 297, 1998.

....protection and information hiding methodologies (see, e.g. 1] 2] 4] and references therein) Relatively little attention, however, has been devoted to the problem of information hiding from an information theoretic perspective. An exception is a recent work by O Sullivan, Moulin, and Ettinger [7], who characterized the highest achievable information rate of watermarking codes for the following system model: A secret message, encoded at rate R, is hidden in a memoryless covertext message within small degradation of quality, symbolized by distortion level D 1 w.r.t. some distortion measure. ....

....D 1 w.r.t. some distortion measure. An active attacker may introduce additional distortion D 2 in attempt to disrupt the watermark. Finally, the resulting data set is analyzed using information shared with the information hider (i.e. the original covertext message) to extract the watermark. In [7], the highest information rate R has been found (as a function of D 1 and D 2 ) for which, even in the presence of optimum attack, there exist block encoders and decoders that guarantee arbitrarily small probability of erroneous watermark decoding. In this paper, we go one step further and provide ....

[Article contains additional citation context not shown here]

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information theoretic analysis of Steganography," Proc. ISIT `98, p. 297, 1998.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information--Theoretic Analysis of Steganography, " Proc. IEEE Int. Symp. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information--Theoretic Analysis of Steganography," Proc. IEEE Int. Symp. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information--Theoretic Analysis of Steganography, " Proc. IEEE Int. Symp. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998.

....(3) becomes tight, and we use the right side of (3) as our design criterion. The actual embedding function is then generated by sampling the optimal distribution p X S . IV. Related Work The role of game theory in watermarking appears in information theoretic analyses by O Sullivan et al. [2, 3, 4], Cohen and Lapidoth [5] and Merhav and Steinberg [6, 7] Optimal detection of known watermarks was studied by Depovere et al. 8] who compared Pe (#, A) to the probability of error P e (#, A) of conventional, suboptimal correlation receivers. Their analysis assumed that no attack takes place ....

....results. V. Gaussian Processes In Secs. VI and VII, we provide two examples where the exact solution to the maxmin problem (1) can be derived. In each example, the signal s(n) n # Z is a stationary, realvalued, Gaussian random process with zero mean and bounded spectral density Ss(f) f # [ 1 2 , 1 2 ]. The host data are a length N segment of a realization of that process; we assume that N is large. We define the level sets for the function Ss(f ) L(z) f : Ss(f) # z , z # 0. 4) If the attack channel is stationary, the distribution p X S that optimizes (3) turns out to ....

[Article contains additional citation context not shown here]

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information-- Theoretic Analysis of Steganography," Proc. IEEE Int. Symp. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998. 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45

....we have [2] min ### max A#A P e (#, A) # min p X S #PX S max A#A P e (p X S , A) 5) where PX S = p X S : E[d(S, X) # E and E[X S] S . 6) 4. RELATED WORK The role of game theory in watermarking appears in the capacity analysis of data hiding systems by O Sullivan et al. [4, 5]. Optimal detection of known watermarks was studied by Depovere et al. 6] who compared P e (#, A) to the probability of error P e (#, A) of conventional, suboptimal correlation receivers. Their analysis assumed that no attack takes place (A is the identity) and that s is a first order Gaussian ....

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information--Theoretic Analysis of Steganography, " Proc. IEEE Int. Symp. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998.

.... to hide data and to remove hidden data [3, 10, 16, 22, 39, 41, 46, 48] However, the informationtheoretic analysis that describes the fundamental limits of any watermarking or data hiding system and can ultimately guide the design of e#cient watermarking algorithms has been emerging only recently [5, 28, 31, 37]. Information theory yields results that are substantially di#erent from those obtained so far in the data hiding literature. A typical approach in this literature is to assume that the data embedding and decoding functions take a particular form, and that a particular type of attack is used [1, ....

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information--Theoretic Analysis of Steganography," Proc. IEEE Int. Symp. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998.

....hiding techniques. Other applications of data hiding include close captioning and embedding of text and audio in images and video. These areas have seen the development of a plethora of algorithms in the last five years [2, 3] but an information theoretic treatment of the problem is just emerging [4, 5, 6, 7, 8, 9]. In particular, a theory has recently been developed to establish the fundamental limits of the watermarking (data hiding) problem depicted in Fig. 1 [4, 5] In this framework, a message M is to be embedded in a length N sequence S N = S 1 , SN ) termed host data set, typically data from ....

....of a plethora of algorithms in the last five years [2, 3] but an information theoretic treatment of the problem is just emerging [4, 5, 6, 7, 8, 9] In particular, a theory has recently been developed to establish the fundamental limits of the watermarking (data hiding) problem depicted in Fig. 1 [4, 5]. In this framework, a message M is to be embedded in a length N sequence S N = S 1 , SN ) termed host data set, typically data from an host image, video, or audio signal. The embedding is done using a cryptographic key. The resulting watermarked data X N = X 1 , XN ) are subject ....

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information--Theoretic Analysis of Steganography," Proc. IEEE Int. Symp. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998.

....compression, signal warping, and addition of noise. Perceptual models for audio, imagery, and video have helped quantify the distortions introduced by information hiding and attack algorithms. Game theoretic aspects of information hiding have been explored for special cases by Ettinger et al. [17, 18]. Cryptographic aspects of information hiding include the use of secret keys to protect the message. It should however be clearly recognized that the functional requirements of cryptography and information hiding as described above are very di#erent, as secrecy of the message is the main objective ....

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information--Theoretic Analysis of Steganography," Proc. IEEE Int. Symp. on Info. Thy, Cambridge, MA, p. 297, Aug. 1998.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger. Information-theoretic analysis of Steganography. Proceedings ISIT 98, 1998.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger. Informationtheoretic analysis of steganography. In Proc. IEEE Symp. on Information Theory, Boston, MA, August 1998.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information-theoretic analysis of steganography," in Proceedings of IEEE International Symposium on Information Theory, 1998, p. 297.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger. Information-theoretic analysis of Steganography. Proceedings ISIT 98, 1998.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, \Information theoretic analysis of steganography, " Proc. ISIT `98, p. 297, 1998.

No context found.

J. A. O'Sullivan, P. Moulin, and J. M. Ettinger, "Information theoretic analysis of steganography," in Proc. IEEE International Symposium on Information Theory, p. 297, August 1998.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC