Patrick Bas

OP-ELM: Optimally Pruned Extreme Learning Machine

In this brief, the optimally pruned extreme learning machine (OP-ELM) methodology is presented. It is based on the original extreme learning machine (ELM) algorithm with additional steps to make it more robust and generic. The whole methodology is presented in detail and then applied to several regression and classification problems. Results for both computational time and accuracy (mean square error) are compared to the original ELM and to three other widely used methodologies: multilayer perceptron (MLP), support vector machine (SVM), and Gaussian process (GP). As the experiments for both regression and classification illustrate, the proposed OP-ELM methodology performs several orders of magnitude faster than the other algorithms used in this brief, except the original ELM. Despite the simplicity and fast performance, the OP-ELM is still able to maintain an accuracy that is comparable to the performance of the SVM. A toolbox for the OP-ELM is publicly available online.

Break our steganographic system: the ins and outs of organizing BOSS

This paper summarizes the first international challenge on steganalysis called BOSS (an acronym for Break Our Steganographic System). We explain the motivations behind the organization of the contest, its rules together with reasons for them, and the steganographic algorithm developed for the contest. Since the image databases created for the contest significantly influenced the development of the contest, they are described in a great detail. Paper also presents detailed analysis of results submitted to the challenge. One of the main difficulty the participants had to deal with was the discrepancy between training and testing source of images - the so-called cover-source mismatch, which forced the participants to design steganalyzers robust w.r.t. a specific source of images. We also point to other practical issues related to designing steganographic systems and give several suggestions for future contests in steganalysis.

TROP-ELM: A double-regularized ELM using LARS and Tikhonov regularization

In this paper an improvement of the optimally pruned extreme learning machine (OP-ELM) in the form of a L2 regularization penalty applied within the OP-ELM is proposed. The OP-ELM originally proposes a wrapper methodology around the extreme learning machine (ELM) meant to reduce the sensitivity of the ELM to irrelevant variables and obtain more parsimonious models thanks to neuron pruning. The proposed modification of the OP-ELM uses a cascade of two regularization penalties: first a L1 penalty to rank the neurons of the hidden layer, followed by a L2 penalty on the regression weights (regression between hidden layer and output layer) for numerical stability and efficient pruning of the neurons. The new methodology is tested against state of the art methods such as support vector machines or Gaussian processes and the original ELM and OP-ELM, on 11 different data sets; it systematically outperforms the OP-ELM (average of 27% better mean square error) and provides more reliable results – in terms of standard deviation of the results – while remaining always less than one order of magnitude slower than the OP-ELM.

Color image watermarking using quaternion Fourier transform

The paper presents a digital color image watermarking scheme using a hypercomplex numbers representation and the quaternion Fourier transform (QFT). Previous color image watermarking methods are first presented and the quaternion representation is then described. In this framework, RGB pixel values are associated with a unique quaternion number having three imaginary parts. The QFT is presented; this transform depends on an arbitrary unit pure quaternion, /spl mu/. The value of /spl mu/ is selected to provide embedding spaces having robustness and/or perceptual properties. In our approach, /spl mu/ is a function of the mean color value of a block and a perceptual component. A watermarking scheme based on the QFT and the quantization index modulation scheme is then presented. This scheme is evaluated for different color image filtering processes (JPEG, blur). The fact that perceptive QFT embedding can offer robustness to luminance filtering techniques is outlined.

Using the fractal code to watermark images

Our paper presents a watermarking scheme based on an insertion of similarities. In the first part different watermarking techniques are presented and classed. In the second part our scheme is described in its spatial and frequential implantations. Finally the different results and perspectives of the work are outlined.

Broken arrows

This paper makes an account of the design and investigations done for the still image watermarking technique used in the 2nd edition of the BOWS challenge. This technique is named “broken arrows” for some reasons given later on, and abbreviated “BA.” This zero-bit algorithm is an implementation of a recent theoretical result by Merhav and Sabbag (2008) with precautions taken with respect to robustness, security, and imperceptibility. A new robustness criterion, based on the nearest border point of a cone, is proposed. The security constraint is taken into account by increasing the diversity of the watermark, sculpturing and randomizing the shape of the detection regions. The imperceptibility and robustness are also provided by adopting proportional embedding in the wavelet domain. The algorithm has been benchmarked using a database of 2000 images. For a probability of false alarm below and a PSNR of 43 dB, the overall robustness regarding various classical image processing seems a promising and strong basis for the challenge.

A new video-object watermarking scheme robust to object manipulation

This paper presents a watermarking scheme for image or video objects. The watcamarking of video objects implies different constraints from raw watermarking methods. The mark has to be detected after object manipulations such as rotations, translations arid VOL modifications. To achieve these requirements, the embedding scheme exploits the shape of the object: a random sequence is transformed to fit the scale and the orientation of the object. The detection of the mark is performed applying an inverse transform and calculating a correlation between the random sequence and the warped object. Our results illustrate the fact that the presented method is robust to object manipulation.This paper presents a watermarking scheme for image or video objects. The watermarking of video objects implies different constraints from raw watermarking methods. The mark has to be detected after object manipulations such as rotations, translations and VOL modifications. To achieve these requirements, the embedding scheme exploits the shape of the object: a random sequence is transformed to fit the scale and the orientation of the object. The detection of the mark is performed applying an inverse transform and calculating a correlation between the random sequence and the warped object. Our results illustrate the fact that the presented method is robust to object manipulation.

A feature selection methodology for steganalysis

This paper presents a methodology to select features before training a classifier based on Support Vector Machines (SVM). In this study 23 features presented in [1] are analysed. A feature ranking is performed using a fast classifier called K-Nearest-Neighbours combined with a forward selection. The result of the feature selection is afterward tested on SVM to select the optimal number of features. This method is tested with the Outguess steganographic software and 14 features are selected while keeping the same classification performances. Results confirm that the selected features are efficient for a wide variety of embedding rates. The same methodology is also applied for Steghide and F5 to see if feature selection is possible on these schemes.

Achieving subspace or key security for WOA using natural or circular watermarking

This paper presents two watermarking schemes that are secure when considering the Watermarked content Only Attack (WOA) framework. The definition of watermark security is first recalled and the distinction between key-security and subspace- security classes for Spread Spectrum (SS) watermarking schemes is presented afterwards. Blind source separation techniques are also recalled as a tool to assess the security of a SS watermarking scheme and prove that classical SS and Improved SS are not secure within the WOA framework. To further illustrate these security issues, we next build on a new watermarking scheme called Natural Watermarking (NW). We prove it to be subspace-secure under specific hypotheses. Natural watermarking does not change the Gaussian natural distributions of the projection of each carriers. Furthermore NWprevents estimations both of the watermark subspace (subspace-security) and the different carriers (key-security). We then extend the nice properties of NW to derive the more general family of Circular Watermarking schemes which are key-secure and may offer a better robustness to AWGN attack than NW. An implementation of CW based on ISS is next proposed and comparison of bit error rates for NW, CW, SS and ISS finally draws some conclusions on the robustness cost to achieve security.

Optimal Transport for Secure Spread-Spectrum Watermarking of Still Images

This paper studies the impact of secure watermark embedding in digital images by proposing a practical implementation of secure spread-spectrum watermarking using distortion optimization. Because strong security properties (key-security and subspace-security) can be achieved using natural watermarking (NW) since this particular embedding lets the distribution of the host and watermarked signals unchanged, we use elements of transportation theory to minimize the global distortion. Next, we apply this new modulation, called transportation NW (TNW), to design a secure watermarking scheme for grayscale images. The TNW uses a multiresolution image decomposition combined with a multiplicative embedding which is taken into account at the distribution level. We show that the distortion solely relies on the variance of the wavelet subbands used during the embedding. In order to maximize a target robustness after JPEG compression, we select different combinations of subbands offering the lowest Bit Error Rates for a target PSNR ranging from 35 to 55 dB and we propose an algorithm to select them. The use of transportation theory also provides an average PSNR gain of 3.6 dB on PSNR with respect to the previous embedding for a set of 2000 images.