In this paper, a spread-spectrum-like discrete cosine transform (DCT) domain watermarking technique for copyright protection of still digital images is analyzed. The DCT is applied in blocks of 8x8 pixels, as in the JPEG algorithm. The watermark can encode information to track illegal misuses. For flexibility purposes, the original image is not necessary during the ownership verification process, so it must be modeled by noise. Two tests are involved in the ownership verification stage: watermark decoding, in which the message carried by the watermark is extracted, and watermark detection, which decides whether a given image contains a watermark generated with a certain key. We apply generalized Gaussian distributions to statistically model the DCT coefficients of the original image and show how the resulting detector structures lead to considerable improvements in performance with respect to the correlation receiver, which has been widely considered in the literature and makes use of the Gaussian noise assumption. As a result of our work, analytical expressions for performance measures, such as the probability of errors in watermark decoding and the probabilities of false alarms and of detection in watermark detection, are derived and contrasted with experimental results.

Growing energy needs are forcing governments to look for alternative resources and ways to better manage the energy grid and load balancing. As a major initiative, many countries including the United Kingdom, United States, and China have already started deploying smart grids. One of the biggest advantages of smart grids compared to traditional energy grids is the ability to remotely read fine-granular measurements from each smart meter, which enables the grid operators to balance load efficiently and offer adapted time-dependent tariffs. However, collecting fine-granular data also poses a serious privacy threat for the citizens as illustrated by the decision of the Dutch Parliament in 2009 that rejects the deployment of smart meters due to privacy considerations. Hence, it is a must to enforce privacy rights without disrupting the smart grid services like billing and data aggregation. Secure signal processing (SSP) aims at protecting the sensitive data by means of encryption and provides tools to process them under encryption, effectively addressing the smart metering privacy problem.

A novel quantization-based data-hiding method, called Rational Dither Modulation (RDM), is presented. This method retains most of the simplicity of the conventional dither modulation (DM) scheme, which is largely vulnerable to amplitude scalings and modifies it in such a way that the result becomes invariant to gain attacks. RDM is based on using a gain-invariant adaptive quantization step-size at both embedder and decoder. This causes the watermarked signal to be asymptotically stationary. Mathematical tools are used to determine the stationary probability density function, which is later used to assess the performance of RDM in Gaussian channels. It is also shown that by increasing the memory of the system, it is possible to asymptotically approach the performance of DM, still keeping invariance against gain attacks. RDM is compared with improved spread-spectrum methods, showing that the former can achieve much higher rates for the same bit error probability. Experimental results confirm the validity of the theoretical analyses given in the paper. Finally, a broader class of methods, that extends gain-invariance to quantization index modulation (QIM) methods, is also presented.

This paper presents both theoretical and practical analyses of the security offered by watermarking and data hiding methods based on spread spectrum. In this context, security is understood as the difficulty of estimating the secret parameters of the embedding function based on the observation of watermarked signals. On the theoretical side, the security is quantified from an information-theoretic point of view by means of the equivocation about the secret parameters. The main results reveal fundamental limits and bounds on security and provide insight into other properties, such as the impact of the embedding parameters, and the tradeoff between robustness and security. On the practical side, workable estimators of the secret parameters are proposed and theoretically analyzed for a variety of scenarios, providing a comparison with previous approaches, and showing that the security of many schemes used in practice can be fairly low.

Face recognition is one of the foremost applications in computer vision, which often involves sensitive signals; privacy concerns have been raised lately and tackled by several recent privacy-preserving face recognition approaches. Those systems either take advantage of information derived from the database templates or require several interaction rounds between client and server, so they cannot address outsourced scenarios. We present a private face verification system that can be executed in the server without interaction, working with encrypted feature vectors for both the templates and the probe face. We achieve this by combining two significant contributions: 1) a novel feature model for Gabor coefficients' magnitude driving a Lloyd-Max quantizer, used for reducing plaintext cardinality with no impact on performance; 2) an extension of a quasi-fully homomorphic encryption able to compute, without interaction, the soft scores of an SVM operating on quantized and encrypted parameters, features and templates. We evaluate the private verification system in terms of time and communication complexity, and in verification accuracy in widely known face databases (XM2VTS, FERET, and LFW). These contributions open the door to completely private and noninteractive outsourcing of face verification.