355 research outputs found
Hermite Polynomials as Provably Good Functions to Watermark White Gaussian Hosts
No full textInternational audienceIn the watermark detection scenario, also known as zero-bit watermarking, a watermark, carrying no hidden message, is inserted in content. The watermark detector checks for the presence of this particular weak signal in content. The article looks at this problem from a classical detection theory point of view, but with side information enabled at the embedding side: the watermark signal is a function of the host content. Our study is twofold. The first issue is to design the best embedding function for a given detection function (a Neyman-Pearson detector structure is assumed). The second issue is to find the best detection function for a given embedding function. This yields two conditions, which are mixed into one 'fundamental' partial differential equation. Solutions of this fundamental equation are heavily dependent on the probability distribution function of the host signals. This conference paper is an extract of~\cite{Furon2006:A-constructive}, where we only look at white gaussian hosts. This gives birth to polynomials solutions known as Hermite polynomial, whose extension is the JANIS watermarking scheme, invented heuristically some years ago
A survey of watermarking security
No full textDigital watermarking studies have always been driven by the improvement of robustness. Most of articles of this field deal with this criterion, presenting more and more impressive experimental assessments. Some key events in this quest are the use of spread spectrum, the invention of resynchronization schemes, the discovery of side information channel, and the formulation of the opponent actions as a game. On the contrary, security received little attention in the watermarking community. This paper presents a comprehensive overview of this recent topic. We list the typical applications which requires a secure watermarking technique. For each context, a threat analysis is purposed. This presentation allows us to illustrate all the certainties the community has on the sub ject, browsing all key papers. The end of the paper is devoted to what remains not clear, intuitions and future studies
Watermarking error exponents in the presence of noise: The case of the dual hypercone detector
No full textInternational audienceThe study of the error exponents of zero-bit watermarking is addressed in the article by Comesana, Merhav, and Barni, under the assumption that the detector relies solely on second order joint empirical statistics of the received signal and the watermark. This restriction leads to the well-known dual hypercone detector, whose score function is the absolute value of the normalized correlation. They derive the false negative error exponent and the optimum embedding rule. However, they only focus on high SNR regime, i.e. the noiseless scenario. This paper extends this theoretical study to the noisy scenario. It introduces a new definition of watermarking robustness based on the false negative error exponent, derives this quantity for the dual hypercone detector, and shows that its performances is almost equal to Costa's lower bound
Watermarking for alternative requirements
No full textWatermarking is a primitive robustly hiding binary messages in host media. However, some applications needs adaptations of this basic primitive. The aim of this chapter is to detail such possible enrichments. However, this chapter is not only a list of less classical watermarking applications. The relationship between cryptography and watermarking is the base supporting this chapter. Cryptography and watermarking tackles the same issue: computer security (but note that watermarking doesn't only target secure applications). The presented enrichments are largely inspired by known functionalities of cryptography. The differences are sometimes quite subtle but important. It would not make sense that watermarking only mimics functionalities of cryptography whose theoretical and practical security levels are assessed for decades. Therefore, this chapter focuses on the interactions between these two technologies. All cryptographic references in this chapter are taken from Menezes et al. (1996). This chapter makes an overview of four different applications: authentication, fingerprinting, watermarking protocols (embedding and detection), and asymmetric watermarking
Constructing new tools for efficient homomorphic encryption
No full textDans notre vie de tous les jours, nous produisons une multitude de données à chaque fois que nous accédons à un service en ligne. Certaines sont partagées volontairement et d'autres à contrecœur. Ces données sont collectées et analysées en clair, ce qui menace la vie privée de l'utilisateur et empêche la collaboration entre entités travaillant sur des données sensibles. Le chiffrement complètement homomorphe (Fully Homomorphic Encryption) apporte une lueur d'espoir en permettant d'effectuer des calculs sur des données chiffrées ce qui permet de les analyser et de les exploiter sans jamais y accéder en clair. Cette thèse se focalise sur TFHE, un récent schéma complètement homomorphe capable de réaliser un bootstrapping en un temps record. Dans celle-ci, nous introduisons une méthode d'optimisation pour sélectionner les degrés de liberté inhérents aux calculs homomorphiques permettant aux profanes d'utiliser TFHE. Nous détaillons une multitude de nouveaux algorithmes homomorphes qui améliorent l'efficacité de TFHE et réduisent voire éliminent les restrictions d'algorithmes connus. Une implémentation efficace de ceux-ci est d'ores et déjà en accès libre.In our everyday life, we leave a trail of data whenever we access online services. Some are given voluntarily and others reluctantly. Those data are collected and analyzed in the clear which leads to major threats on the user's privacy and prevents collaborations between entities working on sensitive data. In this context, Fully Homomorphic Encryption brings a new hope by enabling computation over encrypted data, which removes the need to access data in the clear to analyze and exploit it. This thesis focuses on TFHE, a recent fully homomorphic encryption scheme able to compute a bootstrapping in record time. We introduce an optimization framework to set the degrees of freedom inherent to homomorphic computations which gives non-experts the ability to use it (more) easily. We describe a plethora of new FHE algorithms which improve significantly the state of the art and limit, (if not remove) existing restrictions. Efficient open source implementations are already accessible
L'illusion du test par groupe
No full textThis report challenges the assumptions usually made in non-adaptive group testing. The test is usually modelled as a probabilistic mechanism prone to false positive and / or false negative errors. However, the models are still too optimistic because the performances of these non ideal tests are assumed to be independent of the size of the groups. Without this condition, the report shows that the promises of group test (a number of tests and a decoding complexity scaling as c logN) do not hold.Ce rapport de recherche présente une investigation sur les hypothèses parfois cachées en test par groupe. Pour un nombre c de malades sur une population de taille N, on dit souvent qu’il suffit de O(c logN) tests pour identifier les malades. Ce résultat est erroné dès que les performances du test s’effondrent avec la taille du groupe
A survey of watermarking security
No full textDigital watermarking studies have always been driven by the improvement of robustness. Most of articles of this field deal with this criterion, presenting more and more impressive experimental assessments. Some key events in this quest are the use of spread spectrum, the invention of resynchronization schemes, the discovery of side information channel, and the formulation of the opponent actions as a game. On the contrary, security received little attention in the watermarking community. This paper presents a comprehensive overview of this recent topic. We list the typical applications which requires a secure watermarking technique. For each context, a threat analysis is purposed. This presentation allows us to illustrate all the certainties the community has on the sub ject, browsing all key papers. The end of the paper is devoted to what remains not clear, intuitions and future studies
Watermarking for alternative requirements
No full textWatermarking is a primitive robustly hiding binary messages in host media. However, some applications needs adaptations of this basic primitive. The aim of this chapter is to detail such possible enrichments. However, this chapter is not only a list of less classical watermarking applications. The relationship between cryptography and watermarking is the base supporting this chapter. Cryptography and watermarking tackles the same issue: computer security (but note that watermarking doesn't only target secure applications). The presented enrichments are largely inspired by known functionalities of cryptography. The differences are sometimes quite subtle but important. It would not make sense that watermarking only mimics functionalities of cryptography whose theoretical and practical security levels are assessed for decades. Therefore, this chapter focuses on the interactions between these two technologies. All cryptographic references in this chapter are taken from Menezes et al. (1996). This chapter makes an overview of four different applications: authentication, fingerprinting, watermarking protocols (embedding and detection), and asymmetric watermarking
L'illusion du test par groupe
No full textThis report challenges the assumptions usually made in non-adaptive group testing. The test is usually modelled as a probabilistic mechanism prone to false positive and / or false negative errors. However, the models are still too optimistic because the performances of these non ideal tests are assumed to be independent of the size of the groups. Without this condition, the report shows that the promises of group test (a number of tests and a decoding complexity scaling as c logN) do not hold.Ce rapport de recherche présente une investigation sur les hypothèses parfois cachées en test par groupe. Pour un nombre c de malades sur une population de taille N, on dit souvent qu’il suffit de O(c logN) tests pour identifier les malades. Ce résultat est erroné dès que les performances du test s’effondrent avec la taille du groupe
Comprendre, apprivoiser et se protéger des exemples adversaires
No full textL'Intelligence Artificielle est une discipline qui a connu un fort essor au cours de ces dernières années, notamment en Vision par Ordinateur où l'application la plus commune est la classification d'image. Aujourd'hui, les réseaux de neurones artificiels profonds sont d’excellents classifieurs inférant ce que représente une image. Des travaux ont cependant rapidement montré qu’ils sont vulnérables aux attaques par évasion, aussi appelés les exemples adverses. Ces exemples sont des images qui pour un humain semblent être une représentation normale d'un objet. Mais le classifieur attaqué ne parviendra pas à prédire correctement ce qu'elles représentent.Cette thèse étudie les mécanismes de création de ces exemples, la raison de leur existence et la vulnérabilité des classifieurs. En particulier, ce travail replace ces exemples adverses dans un contexte réaliste. Premièrement, il propose des attaques rapides même sur des grandes images avec un fort taux de succès et une distortion imperceptible ou indétectable.Deuxièmement, il ajoute la contrainte que les exemples adversaires sont avant tout des images, c’est à dire des signaux quantifiés dans le domaine spatial (format PNG) ou dans le domaine DCT (format JPEG).Artificial Intelligence is nowadays one of the most essential disciplines of computer science. These algorithms perform particularly well on Computer Vision tasks, especially classification. A classifier infers what an image represents. Nowadays Deep Neural Networks are largely used for these problems. These neural networks first undergo a training phase during which they are given many examples. These images are accompanied by labels: information on what the image represents. However, it was quickly found that the same logic used during the training phase could be used maliciously. This is the creation of Adversarial Examples through an Evasion Attack.Such examples are seemingly normal images. A human understands what it represents as if it was not manipulated. But the attacked classifier will make an incorrect prediction. In this manuscript, we study the creation of such examples, the reason for their existence, and the underlying vulnerability of classifiers. In particular, we study these examples in a realistic context. First, attacks are optimized (high success rate and low distortion). Second, we add the constraint that adversarial examples should be images. We thus work on spatially-quantized (PNG) or DCT-quantized images (JPEG)
- …
