HAL Portal IOGS (nstitut d'Optique Graduate School)
Not a member yet
12589 research outputs found
Sort by
Efficient Adaptive Multi-Level Privilege Partitioning With RTrustSoC
No full textInternational audienceIn recent years, heterogeneous SoCs (comprised of multiple processor cores and programmable logic) have greatly progressed both complexity and performance. From a security point of view, this leads to an expansion of the attack surface exposed to adversaries. To address this issue, in this article, we propose a novel heterogeneous SoC architecture called RTrustSoC. Our proposal includes an innovative fully-reconfigurable post-deployment strategy for partitioning the SoC architecture into multiple exclusion levels (worlds) with customizable degrees of privilege. We aim to provide SoC designers with fine control over the security of the system by segregating trusted hardware components from third-party IPs with ``on-demand'' hardware isolation. Therefore, we expect that an RTrustSoC instance could evolve from a multi-world SoC to a fully trusted platform as IPs progressively develop. RTrustSoC also proposes a dynamic reconfigurable penalty system to monitor the third-party IPs and take measures in case of a detected abnormal behavior. Our experimental testing on an AMD-Xilinx Zynq-7000 SoC-FPGA showed the penalty of the proposed isolation strategy to be small, up to 1% in LUT and 0.7% Flip Flop utilization, thus enabling to an efficient security solution. RTrustSoC introduces a novel design paradigm, evolving from the binary notion of security (trusted vs untrusted) into a flexible set of worlds that can be adapted to any scenario. We demonstrate a real case scenario of RTrustSoC use on time-based cache memory attacks with implementation results
Optimisation d'un système d'imagerie catoptrique freeform à l'aide de FORMIDABLE : une librairie de conception optique avec des capacités d'optimisation NURBS et de tracé de rayon différentiable
No full textInternational audienc
Lâchez les rênes, laissez vous guider : Installation facilitée et sécurisée d'eLabFTW
No full textInternational audienceLe cahier de laboratoire électronique fait évoluer nos pratiques dans le monde de la recherche. Le CNRS a choisi la solution Open Source eLabFTW avec deux offres (SaaS et On Premise), afin de répondre aux critères de traçabilité, sécurité, confidentialité et pérennité des résultats de recherche. Ce projet a été récompensé par un Cristal collectif en 2024.Passer du papier au numérique peut sembler complexe à mettre en place, c'est pour cette raison que nous proposons de vous guider. Notre équipe de 7 membres, issus de différentes unités de recherche de la région bordelaise, a travaillé ardemment sur l'offre On Premise pour démystifier et faciliter l'installation d'eLabFTW.Nous avons rassemblé toutes les informations et sources disponibles pour vous proposer un modèle d'installation simple et accessible, basé sur la technologie des images Docker, complété par une documentation en ligne dédiée. Nous avons pris en compte les défis techniques demandés par le CNRS pour la mise en place d'eLabFTW, tels que la configuration des VLAN, l'installation des serveurs virtuels, la génération de certificats et bien plus encore.Nous avons également exploré d'autres problématiques, telles que les mises à jour d'eLabFTW et la mise en place de sauvegardes chiffrées. Notre travail a été enrichi par les retours d'expérience de plusieurs ASR. Nous sommes convaincus que notre contribution sera utile à l'ensemble de la communauté ESR et sommes impatients de la partager avec vous. Notre approche collaborative a permis de rendre l'installation d'eLabFTW simple et abordable pour tous
SWLN ou comment déployer des règles de sécurité de SWMB pour Windows dans son unité
No full textInternational audienceSecurity is at the heart of our concerns, and workstations are often the weak links in our information systems. No matter how hard you try to implement security measures and pay lip service to the issue, you can't always control your users' ardor. The SWMB/SWLN solution is here to give you some peace of mind. Developed by colleagues who share your concerns about IT security, SWLN is an open-source tool for strengthening the protection of your IT assets.Combined with the SWMB (Secure Windows Mode Batch) core, the SWLN (Secure Windows Local Network) template enables you to quickly deploy the adjustments needed to secure your workstations in line with ANSSI recommendations for optimum security. Whether your systems run on Windows 10 or 11, in an AD or non-AD environment, SWLN can be customized to suit your security policy and the specific needs of each user (workstation). It's a template that's easy to use and extend. You can enable or disable options, modify values. In short, you're in control. By keeping a record of your files, you'll also be able to track who pushed which rule.Come along to our presentation at JRES and discover all the advantages of this free community template! Join the community of SWMB/SWLN users and say goodbye to some of your IT security nightmares.La sécurité est au cœur de nos préoccupations, et les postes de travail sont souvent des maillons faibles de nos systèmes d'information. Vous avez beau essayer de mettre en place des mesures de sécurité et prodiguer de bonnes paroles, vous ne pouvez pas toujours contrôler les ardeurs de vos utilisateurs. La solution basée sur le couple SWMB/SWLN est là pour vous apporter un peu de sérénité. Développé par des collègues qui partagent vos préoccupations en matière de sécurité informatique, SWLN est un outil libre permettant de renforcer la protection de votre parc informatique.Associé au coeur SWMB (Secure Windows Mode Batch), le template SWLN (Secure Windows Local Network) permet de déployer rapidement les ajustements nécessaires pour sécuriser vos postes en respectant les recommandations de l'ANSSI pour une sécurité optimale. Que vos systèmes tournent sous Windows 10 ou 11, qu'ils soient dans un environnement AD ou hors AD, SWLN est personnalisable afin de s'adapter à votre politique de sécurité et aux besoins spécifiques de chaque utilisateur (poste de travail). C'est un template facile à utiliser et à étendre. Vous pouvez activer ou désactiver des options, modifier des valeurs. Bref, vous êtes aux commandes. En historisant vos fichiers, vous bénéficiez en plus d'un suivi permettant de savoir qui a poussé telle ou telle règle.Venez assister à notre présentation lors des JRES et découvrez tous les avantages de ce template communautaire libre ! Venez rejoindre la communauté des utilisateurs du couple SWMB/SWLN et dire adieu à certains de vos cauchemars de sécurité informatique
Deciphering the complexity behind laser-induced selforganized nanopatterns
No full textInternational audienceWhen subjected to a series of ultrashort laser pulses, an initially smooth surface undergoes a progressive transformation into dissipative structures, empowering it to regulate energy flux and optimize entropy production rates. This phenomenon manifests as the spontaneous emergence of periodic patterns at various scales, with diverse shapes and varying aspect ratios. Consequently, numerous 2D patterns (stripes, labyrinthines, dots, triangles, hexagonal cavities, etc.) have recently been observed at various scales [1]. The theoretical challenge lies in developing an effective model with symmetry breaking, scale invariance, stochasticity, and nonlinear properties to replicate dissipative structures. Describing pattern growth requires nonlinear dynamics under far-from-equilibrium conditions, for which classical equations (Maxwell, Navier-Stokes, Fourier…) demand unknown transient material properties [2]. We have implemented a stochastic Swift-Hohenberg model that replicates hydrodynamic fluctuations near the convective instability threshold, inherent in laserinduced self-organized nanopatterns at the nanoscale. We will show that a deep convolutional network can learn the complexity of patterns, linking model coefficients to experimental parameters to design specific morphologies [3]. The model accurately predicts patterns, identifying laser parameter regions and potentially anticipating physics complexity evolution.[1] A. Nakhoul, and J.P. Colombier, “Beyond the Microscale—Advances in Surface Nanopatterning by Laser-Driven Self-Organization”, Laser & Photonics Reviews, In press (2024).[2] A. Rudenko and J.P. Colombier, “How light drives material periodic patterns down to the nanoscale”, Ultrafast Laser Nanostructuring: The Pursuit of Extreme Scales, 209-255 (2023). [3] E. Brandao, A. Nakhoul S. Duffner, R. Emonet, F. Garrelie, A. Habrard, F. Jacquenet, F. Pigeon, M. Sebban. & J.P. Colombier, “Learning complexity to guide light-induced self-organized nanopatterns”, Physical Review Letters, 130, 226201 (2023)
Ultrafast laser-irradiated silica: from excited-state dynamics to nanostructuring
No full textInternational audienceUnderstanding the response of wide-bandgap materials such as silicon dioxide (fused silica, α-quartz) to light is crucial for achieving precise control and manipulation in ultrafast laser volume structuring at the nanoscale. Surpassing optical resolution, this scale of achievement necessitates manipulating light energy around bulk inhomogeneities while precisely orchestrating selective thermodynamic pathways for absorption confinement, crystal-amorphous transformation and rapid energy quenching within nanometer lengths. We propose multiphysics calculations to elucidate the intricate interplay between electronic structure alterations and structural/hydrodynamical relaxation mechanisms under extreme nonequilibrium conditions. In particular, ab initio calculations reveal a narrowing of the bandgap by several eV and a loss of cohesion within an ultrashort timescale. This approach enables precise control and manipulation, optimizing processing parameters, and exploring novel aspects of solid relaxation induced by intense photoexcitation. Ultrafast excitation from a high-power ultrashort laser pulse leads to a surge in transient electron density within the conduction band, causing band distortion and significant bandgap renormalization. Using Density Functional Theory (DFT) within the GW approximation, including Time-Dependent DFT and molecular dynamics, we investigate the evolution of the bandgap across various excitation levels, revealing shifts of several electronvolts within ultrashort timescales. Spatial redistribution of excited charges weakens silica bonds, prompting crystal structure reorganization and further impacts on optical and thermal electronic properties. At later times, complex modulated absorption induces self-organized structures and structural rearrangement. Electromagnetic coupled with hydrodynamics simulations of surface and volume self-organization unveil the dynamic interplay between light and matter evolution. The origin of this effect lies in 3D self-organization, with nanogratings oriented by polarization exhibiting a birefringence function, indicating reshaping of nanoscale plasmas under strong fields and potential resonance development. These findings advance our understanding of ultrafast laser-material interactions, facilitating optimized laser-based processing techniques
ChromaPose: Robustness of 2D Pose Estimation Under Different Color Illuminations
No full textInternational audienc
Length Independent PAC-Bayes Bounds for Simple RNNs
No full textInternational audienceWhile the practical interest of Recurrent Neural Networks (RNNs) is attested, much remains to be done to develop a thorough theoretical understanding of their abilities, particularly in what concerns their learning capacities. A powerful framework to tackle this question is the one of PAC-Bayes theory, which allows one to derive bounds providing guarantees on the expected performance of learning models on unseen data. In this paper, we provide an extensive study on the conditions leading to PAC-Bayes bounds for non-linear RNNs that are independent of the length of the data. The derivation of our results relies on a perturbation analysis on the weights of the network. We prove bounds that hold for β-saturated and DS β-saturated SRNs, classes of RNNs we introduce to formalize saturation regimes of RNNs. The first regime corresponds to the case where the values of the hidden state of the SRN are always close to the boundaries of the activation functions. The second one, closely related to practical observations, only requires that it happens at least once in each component of the hidden state on a sliding window of a given size
Entropy Computation for Oscillator-based Physical Random Number Generators
No full textInternational audienceIn this paper, we provide a complete set of algorithms aimed at the design and security evaluation of oscillator-based True Random Number Generators (TRNG). While depending on some TRNG design assumptions, the proposed algorithms use as inputs the statistical parameters of the underlying random physical process such as the clock jitter originating from the thermal noise and give a lower bound of the entropy rate of the generated bit stream as output. We describe the general structure of a TRNG composed of multiple free-running oscillators and samplers, the outputs of which are post-processed by an entropy conditioner. Depending on the specification of the entropy conditioner, which can usually be any Boolean function, we describe several algorithmic optimizations. We then explain how to compute and efficiently manage the entropy rate at the output of such a post-processing block and at the output of the generator as a whole
A General Framework for the Practical Disintegration of PAC-Bayesian Bounds
No full textInternational audiencePAC-Bayesian bounds are known to be tight and informative when studying the generalization ability of randomized classifiers. However, they require a loose and costly derandomization step when applied to some families of deterministic models such as neural networks. As an alternative to this step, we introduce new PAC-Bayesian generalization bounds that have the originality to provide disintegrated bounds, i.e., they give guarantees over one single hypothesis instead of the usual averaged analysis. Our bounds are easily optimizable and can be used to design learning algorithms. We illustrate this behavior on neural networks, and we show a significant practical improvement over the state-of-the-art framework