HAL Portal IOGS (nstitut d'Optique Graduate School)
Not a member yet
12589 research outputs found
Sort by
Federated Representation Learning for Encrypted Application Type Classification in beyond 5G RAN
No full textInternational audienceMobile application classification is essential for advanced network management and application-based QoS policy enforcement in future, AI-enhanced, beyond 5G and 6G mobile networks. This article proposes to use AI methods to categorize applications as functional types (e.g., Video, Audio, Browsing) despite encryption and limited labeled data. We tackle these challenges through unsupervised representation learning, which maximizes the use of abundant unlabeled data in mobile networks. Due to the distributed nature of beyond 5G and 6G networks, we use this method in federated learning scenarios and compare it to the centralized ones. Our findings highlight that unsupervised learning improves model performance, especially with scarce labeled data. Additionally, federated learning provides effective results as compared to centralized methods.</div
Practical Persistent Fault Attacks on AES with Instruction Skip
No full textInternational audiencePersistent Fault Attacks (PFA) have emerged as an active research area in embedded cryptography. This attack exploits faults in one or multiple constants stored in memory, typically targeting S-box elements. In the literature, such persistent faults primarily induced by bit flips in storage, often achieved through laser fault injection techniques. In this paper, we demonstrate that persistent faults can also be induced through instruction skips, which can easily be achieved with almost any fault injection methods (e.g., voltage/clock glitching, electromagnetism). Specifically, we target AES implementations that dynamically generate the S-box table at runtime, during the initialization phase, before executing the first AES operation. We illustrate this with an attack on the AES implementation in the MbedTLS library, where a clock glitch is inserted during the S-box generation. Secondly, we introduce, to our knowledge, the first PFA that targets a constant other than the S-box elements. We show that faulting a round constant involved in the AES key schedule is sufficient to recover the key by a differential analysis. Compared to previous PFAs that rely on statistical analysis requiring hundreds to thousands of ciphertexts, our approach needs only three correct-faulty ciphertexts pairs. We showcase this attack with an experiment on the MbedTLS AES implementation, using a clock glitch in the round constant generation
Influence of Manufacturing Parameters on the Steady-State Radiation Response of Radiation-Sensitive Optical Fibers
No full textInternational audienc
Analysis of laser-induced damage threshold of circular grating waveguide structures exposed to sub-picosecond laser radiation centered at a wavelength of 1030 nm
No full textInternational audienceThe laser-induced damage threshold of a grating waveguide output coupler (GWOC) exposed to laser radiation at a wavelength of 1030 nm and with a pulse duration of 500 fs was investigated. The GWOC is a combination of a sub-wavelength circular grating and a partial reflector based on a Nb 2 O 5 and SiO 2 multilayer sequence. It was designed to be used as an output coupler of a thin-disk laser cavity for the generation of beams with radial polarization. The results revealed a laser-induced damage threshold (LIDT) fluence of 0.36 J/cm² for single-pulse tests and 0.26 J/cm² for multiple-pulse conditions with up to 1000 shots. These threshold values are comparable to those of an unstructured output coupler with Nb 2 O 5 and SiO 2 coating layers, highlighting the minor influence of the grating on the LIDT
ATOMISTIC-CONTINUUM MODELING OF LASER-INDUCED PHASE TRANSITIONS INSILICON: MELTING, ABLATION, SOLIDIFICATION, AND AMORPHIZATION
No full textInternational audienceModern semiconductor applications demand precise laser processing at the nanometerscale, requiring a detailed understanding of phase transition and structural modificationmechanisms. Accurate control over laser-induced processes in semiconductors isessential for generating predesigned surface structures and modifying the surfaceproperties.In this study, we present a numerical investigation of non-equilibrium laser-inducedphase transitions in silicon (Si) using a hybrid atomistic-continuum model [1]. The modelcombines the strengths of Molecular Dynamics (MD) simulations for atomistic-scaledescriptions of non-equilibrium phase transitions with a continuum approach to accountfor the effect of laser-generated free carriers. As compared to the ordinary continuum orMD approaches, this advanced framework, therefore, captures the kinetics of melting and ablation phenomena on one hand, and generation and diffusion of the electron-hole pairs, thermal diffusion, and the electron-phonon coupling processes during laser energy deposition on the other hand. We applied the model to determine the melting depth as a function of fluence for a 100 fs laser pulse at 800 nm. The results show that the standalone continuum approach underestimates the melting threshold as compared to the hybrid atomistic-continuum model by 46% originating from the detailed description of the melting kinetics. Additionally, we explored the effect of crystal orientation on melting dynamics and compared the results with the corresponding experimental measurement.Finally, the MD model is used to identify the conditions leading to the amorphization ofthe Si surface and corresponding cooling rates are referred to the experimentalconditions. These findings provide valuable insights into experimental observations of Sisurface structuring induced by ultrashort laser pulses. [1] V.P. Lipp, B. Rethfeld, M.E. Garcia, and D.S. Ivanov, “Atomistic-Continuum Modeling of Short Laser Pulse Melting of Si Targets”, Phys Rev B 90, 245306 (2014)
Méthode de quadrature pour les PINNs fondée théoriquement sur la hessienne des résiduels
No full textInternational audiencePhysics-informed Neural Networks (PINNs) have emerged as an efficient way to learn surrogate neural solvers of PDEs by embedding the physical model in the loss function and minimizing its residuals using automatic differentiation at so-called collocation points. Originally uniformly sampled, the choice of the latter has been the subject of recent advances leading to adaptive sampling refinements. In this paper, we propose a new quadrature method for approximating definite integrals based on the hessian of the considered function that we leverage to guide the selection of the collocation points during the training process of PINNs.Les réseaux de neurones informés par la physique (PINNs) sont apparus comme un moyen efficace d'apprendre des solveurs d'EDPs en incorporant le modèle physique dans la fonction de perte et en minimisant ses résiduels par différenciation automatique à des points dits de collocation. Originellement sélectionnés de manière uniforme, le choix de ces derniers a fait l'objet d'avancées récentes en échantillonnage adaptatif. Nous proposons ici une nouvelle méthode de quadrature pour l'approximation d'intégrale basée sur la hessienne de la fonction considérée, pour laquelle nous dérivons une borne d'erreur d'approximation. Nous exploitons cette information de second ordre pour guider la sélection des points de collocation durant l'apprentissage des PINNs
Fair Text Classification via Transferable Representations
No full textGroup fairness is a central research topic in text classification, where reaching fair treatment between sensitive groups (e.g., women and men) remains an open challenge. We propose an approach that extends the use of the Wasserstein Dependency Measure for learning unbiased neural text classifiers. Given the challenge of distinguishing fair from unfair information in a text encoder, we draw inspiration from adversarial training by inducing independence between representations learned for the target label and those for a sensitive attribute. We further show that Domain Adaptation can be efficiently leveraged to remove the need for access to the sensitive attributes in the dataset we cure. We provide both theoretical and empirical evidence that our approach is well-founded
Techniques avancées d'attaques par injection de fautes sur circuits intégrés
No full textThe security of integrated circuits is evaluated through the implementation of attacks that exploit their inherent hardware vulnerabilities. Fault injection attacks represent a technique that is commonly employed for this purpose. These techniques permit an attacker to alter the nominal operation of the component in order to obtain confidential information. The principal techniques for localised fault injection are laser and electromagnetic injection. Recently, pioneering work has demonstrated that X-rays can also modify the behaviour of a circuit. The objective of this thesis is to assess the potential of fault attacks. This is achieved by advancing the existing state of the art, with a particular focus on laser and X-ray fault injection attacks on Flash memories of microcontrollers dedicated to the IoT. The aim of this study is twofold : firstly, to highlight the threat posed by these attacks and secondly, to gain insight into the associated mechanisms. These steps are crucial for the development of effective countermeasures. Firstly, after outlining the limitations of single-spot laser benches, we present a detailed analysis of the significant advantages offered by a new multispot laser bench, particularly in terms of spatial and temporal capacity. The study goes on to describe a number of concrete examples of scenarios that are now achievable, and also carries out a theoretical exploration of the new possibilities offered by the laser bench. Secondly, we propose the utilisation of the thermal effect of an infrared laser bench for the injection of persistent faults into the Flash memory of unpowered components. This novel attack vector gives rise to the delineation of a comprehensive new fault model, encompassing both the physical and application levels. The outcomes obtained facilitate the reverse engineering of the Flash memory of the targeted component and the extraction of the encryption key for a software implementation of the AES encryption algorithm. The final section of the thesis describes the use of unfocused X-ray sources for the injection of faults into the Flash memories of both powered and unpowered components. Furthermore, the thermal and temporal recovery phenomena are also characterised. The design and characterisation of masks that enable the focused injection of faults are demonstrated.La sécurité physique des circuits intégrés est souvent évaluée en menant des attaques qui exploitent leurs vulnérabilités matérielles. Les attaques par injection de fautes sont une technique couramment utilisée dans cet objectif d’évaluation. Elles permettent à un attaquant d’altérer le fonctionnement nominal du composant afin d’obtenir des informations confidentielles. Les techniques principales d’injection de fautes localisées sont les injections laser et électromagnétique. Plus récemment, des travaux pionniers ont montrés que les rayons X pouvaient également modifier le comportement d’un circuit. L’objectif de cette thèse est d’évaluer le potentiel des attaques en fautes. Cela se fait en améliorant l’état de l’art existant, notamment sur les attaques par injection de fautes laser et rayons X, sur des mémoires Flash de microcontrôleurs dédiés à des applications IoT. La finalité de cette étude est de contribuer à la prise en compte de la menace que constituent ces attaques mais également de comprendre les phénomènes associés. Ces points constituent les premiers pas en vue de la conception de contre-mesures adaptées. Premièrement, après une description des limitations des bancs laser monospot, nous caractérisons les avantages significatifs, notamment d’un point de vue spatial et temporel, apportés par un nouveau banc laser multispot. Des exemples concrets de scénarios désormais atteignables sont décrits et une exploration théorique des nouvelles possibilités offertes par le banc laser est également réalisée. Deuxièmement, nous mettons en lumière la possibilité d’utiliser l’effet thermique d’un banc laser infrarouge afin d’injecter des fautes persistantes au sein de mémoires Flash de composants non alimentés. Ce nouveau vecteur d’attaque aboutit à la description d’un nouveau modèle de faute complet allant du niveau physique au niveau applicatif. Les résultats obtenus nous permettent d’une part, de réaliser l’ingénierie inverse de la mémoire Flash du composant ciblé et d’autre part, de retrouver la clé de chiffrement d’une implémentation logicielle de l’algorithme de chiffrement AES. Pour finir, l’utilisation de sources non focalisées de rayons X est décrite dans le but d’injecter des fautes dans des mémoires Flash de composants alimentés et non alimentés. Les phénomènes de récupération thermique et temporelle sont également caractérisés. La conception et la caractérisation de masques permettant, dans une certaine mesure, de focaliser l’injection de fautes est mise en pratique
Combined theoretical and experimental study of luminescence in Gd<sup>3+</sup> -doped silica glass under ionizing radiation
No full textInternational audienceIn the ionizing radiation dosimetry domain, there is still an important demand for a compact, reliable, and accurate sensor. Currently, among the dosimeters based on optical techniques, the fibered systems are the most promising solutions. Optical dosimeters usually exploit one of the physical processes, which is radioluminescence (RL), thermoluminescence (TL) or radiation-induced attenuation (RIA). In this paper, we present, to our knowledge, a new approach to measuring the dose of ionizing radiations that combines radioluminescence, phosphorescence, and thermoluminescence processes. To our knowledge, the integrated signals originating from these three processes have been collected at different temperatures through an optical fiber spliced to a Gd3+-doped silica glass rod for the first time. Using this fibered setup, the experimental study shows that the integrated response, originating from the three processes, is linear versus dose at least up to 300 Gy and temperature-independent between 153 K and 353 K. These experimental results have been explained and reproduced by a theoretical model, which is also able to predict the dosimeter response under various conditions. Moreover, as another novelty, to our knowledge, simultaneous optically stimulated luminescence (OSL) and RL measurements allowed the achievement of a real-time and temperature-independent dosimeter
Three-body interactions in Rabi-coupled Bose gases: a perturbative approach
No full textInternational audienceIn ultracold atomic gases, radio-frequency coupling between two spin states can lead to atoms being in a stable coherent superposition of the two states (dressed states). When the two-body interactions (scattering lengths) are different between the two states, it offers the possibility not only to control the interatomic interaction but also to modify the equation of states with emerging three-body terms with different signs and scalings with the parameters. We derive these terms in a unified perturbation framework and show that they correspond to genuine three-body processes found at different orders in perturbation theory. Our work justifies the introduction of contact three-body interactions in dressed-state systems when the spin degree of freedom is integrated out