Perturbed quantization steganography with wet paper codes

Abstract

In this paper, we introduce a new approach to passive-warden steganography in which the sender embeds the secret message into a certain subset of the cover object without having to share the selection channel with the recipient. An appropriate informationtheoretical model for this communication is writing in memory with (a large number of) defective cells [1]. We describe a simple variable-rate random linear code for this channel (the "wet paper" code) and use it to develop a new steganographic methodology for digital media files -Perturbed Quantization. In Perturbed Quantization, the sender hides data while processing the cover object with an information-reducing operation, such as lossy compression, downsampling, A/D conversion, etc. The sender uses the cover object before processing as side information to confine the embedding changes to those elements of the processed cover object whose values are the most "uncertain". This informed-sender embedding and uninformed-recipient message extraction improves steganographic security because an attacker cannot easily determine from the processed stego object the location of embedding changes. Heuristic is presented and supported by blind steganalysis [2] that a specific case of Perturbed Quantization for JPEG images is significantly less detectable than current JPEG steganographic methods. Categories and Subject Descriptors MOTIVATIONThe primary goal of steganography is to build a statistically undetectable communication channel (the famous Prisoner Problem [3]). In order to embed a secret message, the sender slightly modifies the cover object to obtain the embedded stego object. In steganography under the passive warden scenario [4,5], the goal is to communicate as many bits as possible without introducing any detectable artifacts into the cover object. Attempts to give a formal definition of the concept of steganographic security can be found in [5][6][7][8]. In practice, a steganographic scheme is considered secure if no existing attack can be modified to build a detector that would be able to distinguish between cover and stego images with a success better than random guessing.One possible measure to improve the security of steganographic schemes for digital media is to embed the message in adaptively selected components of the cover object [9-11], such as noisy areas or segments with a complex texture. However, if the adaptive selection rule is public or only "weakly dependent on a key", the attacker can apply the same rule and start building an attack. It is then a valid question whether the adaptive selection improves steganographic security at all. A good example of a scheme where adaptive pixel selection in fact decreased its security is the recent surprising result of Westfeld [12].This problem with adaptive steganography could be remedied if the selection rule was determined from some side information available only to the sender but in principle unavailable to the recipient (and thus any attacker). For example, imagine the situation when the send...

Towards Lower Bounds on Embedding Distortion in Information Hiding

“…An early quantitative technique for steganalysis was designed by Westfeld and Pfitzmann [26]. This research prompted interest in both improving statistical detection techniques [13,15] as well as building new steganographic methods that would be difficult to detect by statistical methods [27,24,16].…”

“…Fridrich et al [16] have proposed an information hiding method that guarantees low distortion rates of stego objects. The method makes use of the JPEG quantization errors by computing all rounding errors of the JPEG coefficients.…”

A Novel Image Steganography in Contourletdomain Using Genetic Algorithm

“…Although modification of properly selected DCT coefficient during embedding process will not cause noticeable visualartifacts, nevertheless they cause detectable statistical degradations [3]. Various steganography methods like YASS [7], MB [8], Outguess [9], Perturbed Quantization (PQ) [10] have been proposed with the purpose of minimizing the statistical artifacts which are produced by modifications of DCT coefficients.…”

Effect of Recompression on Attacking JPEG Steganographic Schemes An Experimental Study

“…There, two types of confusion tests were considered: 1) the original JPEG test image versus the recompressed image with the quality factor estimated from the original image. (this occurs with OutGuess and F5); 2) the original test image versus the recompressed image with a quantization step as twice as the estimated original quantization step (this happens in perturbed quantization (PQ) steganography [10]). …”