Portail HAL des publications du LIRMM
Not a member yet
13279 research outputs found
Sort by
Scoop: An Optimization Algorithm for Profiling Attacks against Higher-Order Masking
No full textInternational audienceIn this paper we provide new theoretical and empirical evidences that gradient-based deep learning profiling attacks (DL-SCA) suffer from masking schemes. This occurs through an initial stall of the learning process: the so-called plateau effect. To understand why, we derive an analytical expression of a DL-SCA model targeting simulated traces which enables us to study an analytical expression of the loss. By studying the loss landscape of this model, we show that not only do the magnitudes of the gradients decrease as the order of masking increases, but the loss landscape also exhibits a prominent saddle point interfering with the optimization process. From these observations, we (1) propose the usage of a second-order optimization algorithm mitigating the impact of low-gradient areas. In addition, we show how to leverage the intrinsic sparsity of valuable information in SCA traces to better pose the DL-SCA problem. To do so, we (2) propose to use the implicit regularization properties of the sparse mirror descent. These propositions are gathered in a new publicly available optimization algorithm, Scoop. Scoop combines second-order derivative of the loss function in the optimization process, with a sparse stochastic mirror descent. We experimentally show that Scoop pushes further the current limitations of DL-SCA against simulated traces, and outperforms the state-of-theart on the ASCADv1 dataset in terms of number of traces required to retrieve the key, perceived information and plateau length. Scoop also performs the first nonworst- case attack on the ASCADv2 dataset. On simulated traces, we show that using Scoop reduces the DL-SCA time complexity by the equivalent of one masking order
Петербургские математики и их открытия (англ). М.:МЦНМО, 2024, рецензия
No full textBook reviewReview of the book (a collection of biographical and scientific notes about St.Petersburg mathematicians) edited by Nikita Kalini
Evaluating podcasts as a tool for OSCE training: a randomized trial using generative AI-powered simulation
No full textInternational audienceIntroduction: Objective Structured Clinical Examinations (OSCEs) are critical for assessing clinical competencies in medical education. While traditional teaching methods remain prevalent, this study introduces an innovative approach by evaluating the effectiveness of an OSCE preparation podcast in improving medical students' OSCE performance using nephrology as a proof of concept. This novel method offers a flexible and accessible format for supplementary learning, potentially revolutionizing medical education.Methods: A mono-centric randomized controlled trial was conducted among 50 fourth-year medical students. Participants were randomly assigned to either the podcast intervention group or a control group. Both groups completed six nephrology-specific OSCE stations on DocSimulator, a generative AI-powered virtual patient platform. Scores from three baseline and three post-intervention OSCE stations were compared. The primary outcome was the change in OSCE scores. Secondary outcomes included interest in nephrology and students' self-reported competence in nephrology-related skills.Results: The baseline OSCE scores did not differ between the two groups (23.8 ± 3.9 vs. 23.3 ± 5.3; p = 0.77). After the intervention, the podcast group demonstrated a significantly higher OSCE score compared to the control group (27.6 ± 3.6 vs. 23.6 ± 5.0; p = 0.002) with a greater improvement in OSCE scores (+ 3.52[0.7,6.5] vs. -1.22[-3,5.5]; p = 0.03). While the podcast did not increase students' intention to specialize in nephrology (4.2% vs. 4.0%; p = 0.99), it significantly improved their confidence in nephrology-related clinical skills (41.7% vs. 16%, p = 0.04). 68% of students in the podcast group found OSCE training podcast useful for their OSCE preparation, and 96% reported they would use it again.Conclusions: The use of an OSCE preparation podcast significantly enhanced students' performance in AI-based simulations and confidence in nephrology clinical competencies. Podcasts represent a valuable supplementary tool for medical education, providing flexibility and supporting diverse learning styles.Trial registration: Not applicable
Graph Embeddings Meet Link Keys Discovery for Entity Matching
No full textInternational audienceEntity Matching (EM) automates the discovery of identity links between entities within different Knowledge Graphs (KGs). Link keys are crucial for EM, serving as rules allowing to identify identity links across different KGs, possibly described using different ontologies. However, the approach for extracting link keys struggles to scale on large KGs. While embedding-based EM methods efficiently handle large KGs they lack explainability. This paper proposes a novel hybrid EM approach to guarantee the scalability link key extraction approach and improve the explainability of embeddingbased EM methods. First, embedding-based EM approaches are used to sample the KGs based on the identity links they generate, thereby reducing the search space to relevant sub-graphs for link key extraction. Second, rules (in the form of link keys) are extracted to explain the generation of identity links by the embedding-based methods. Experimental results demonstrate that the proposed approach allows link key extraction to scale on large KGs, preserving the quality of the extracted link keys. Additionally, it shows that link keys can improve the explainability of the identity links generated by embedding-methods, allowing for the regeneration of 77% of the identity links produced for a specific EM task, thereby providing an approximation of the reasons behind their generation
Proof-theoretic aspects of the logic of scope
No full textInternational audienceIn this paper, we present a proof-theoretic analysis of the logic of scope introduced by Barker and Shan. We notably introduce a novel calculus of proof nets and prove it is sound and complete with respect to the sequent calculus for the logic. We study decidability and complexity of the logic using this new calculus, proving a new upper bound for complexity of the logic (showing it is in NP) and a new lower bound for the class of formal language generated by the formalism (mildly context-sensitive languages extended with a permutation closure operation). Finally, thanks to this new calculus, we present a novel comparison between the logic of scope and the hybrid type-logical grammars of Kubota and Levine. We show there is an unexpected convergence of the natural language analyses proposed in the two formalisms. In addition to studying the proof-theoretic properties of the logic of scope, we greatly extends its linguistic coverage
Colorectal Cancer Tumor Grade Segmentation in Digital Histopathology Images: from GIGA to Mini Challenge
No full textInternational audienceColorectal cancer (CRC) is the third most diagnosed cancer and the second leading cause of cancer-related death worldwide. Accurate histopathological grading of CRC is essential for prognosis and treatment planning but remains a subjective process prone to observer variability and limited by global shortages of trained pathologists. To promote automated and standardized solutions, we organized the ICIP Grand Challenge on Colorectal Cancer Tumor Grading and Segmentation using the publicly available METU CCTGS dataset. The dataset comprises 103 whole-slide images with expert pixellevel annotations for five tissue classes. Participants submitted segmentation masks via Codalab, evaluated using metrics such as macro F-score and mIoU. Among 39 participating teams, six outperformed the Swin Transformer baseline (62.92 F-score). This paper presents an overview of the challenge, dataset, and the top-performing methods
Enhancing Humanoid Robot Stability on Uneven Terrain with Online Foothold Planning Based on Ground Scanning
No full textInternational audienceThis article presents an algorithm for online estimation of foot position and orientation based on the ground profile, designed to improve humanoid stability during locomotion and stationary poses. The method leverages foot-mounted terrain sensing to determine the optimal foothold and orientation that maximize contact surface on uneven terrain. To address hardware limitations, we also integrate compliant feet that enhance adaptability to rough ground. We evaluated the approach through terrain estimation tests and walking experiments in both simulation and on a real humanoid robot with flat and compliant feet. Results show that the algorithm increases the number of successful steps, improves stability, reduces peak ground reaction forces, and lowers ankle angle variations, by enhancing locomotion safety
Body Bias Injection on the FLASH Memory Accelerator of a 32-Bit Microcontroller
No full textInternational audienceProgram flow attacks involve disrupting the flow of instruction execution in microcontrollers (MCUs), thereby threatening their operation. While traditional studies focus on program counter or instruction corruptions within pipelines, little attention has been paid to the stages between FLASH memory and the CPU, such as memory accelerators. Body Bias Injection (BBI) is a fault injection technique in which a voltage pulse is applied to the backside of an integrated circuit, i.e. its substrate, causing localized disruptions in the power network. Despite its proven effectiveness in inducing transient faults, to the best of our knowledge, there is no information on its impact on MCU program flow. Within this context, this paper demonstrates that BBI can efficiently disrupt MCU program flow, causing entire instruction lines to be skipped or repeated. It also shows that the most sensitive part of the MCUs against BBI is likely to be the memory accelerator rather than the processor itself
Model-based thermal drift compensation for high-precision hexapod robot actuators
No full textInternational audienceThermal expansion is a significant source of positioning error in high-precision hexapod robots (Gough-Stewart platforms). Any variation in the temperature of the hexapod's parts induces expansion, which alters their kinematic model and reduces the robot's accuracy and repeatability. These variations may arise from internal heat sources (such as motors, encoders, and electronics) or from environmental changes. In this study, a method is proposed to anticipate and therefore correct the thermal drift of one of the hexapod precision electro-mechanical actuators. This method is based on determining a model that links the expansion state of the actuator at any given moment to the temperature of some well-chosen points on its surface. This model was initially developed theoretically. Its coefficients were then adjusted experimentally on a specific test-bench, based on a rigorous measurement campaign of actuator expansion using a high-precision interferometric measurement system. Experimental validation demonstrates a reduction of thermally induced expansion by more than 80%. This paves the way for thermal drift correction across the entire robot or similar robotics parts
SRAM Periphery Testing using the Cell-Aware Test Methodology
No full textInternational audienceTesting memory circuits is crucial for ensuring the quality and reliability of System-on-Chip (SoC) designs, especially as shrinking technology nodes increase susceptibility to nanometer-scale defects. This paper introduces an enhanced methodology for memory testing, leveraging the Cell-Aware (CA) test concept. Building on prior work for SRAM array testing [1], we extend the CA methodology to include periphery testing by generating, for the first time, CA models for each memory Input-Output (I/O) element, covering key components such as address decoders, write drivers, and sense amplifiers. We present results from testing these periphery components using the CA methodology. Additionally, we compare existing SRAM testing techniques with our CA methodology for the decoder and I/O circuitry. To ensure a fair comparison, we selected minimal March tests designed to detect functional faults in peripheral circuits, aligning with the fault models targeted by our approach. A quantitative analysis of fault coverage demonstrates the effectiveness of our methodology compared to March algorithms, particularly in terms of test complexity