Publications

2025

• Time-resolved second-order autocorrelation function of parametric down-conversion
  
  • Horoshko Dmitri
  • Srivastava Shivang
  • Sośnicki Filip
  • Mikołajczyk Michał
  • Karpiński Michał
  • Brecht Benjamin
  • Kolobov Mikhail
  Physical Review A, American Physical Society, 2025, 112 (2), pp.023703-1:023703-13. We study a possibility of measuring the time-resolved second-order autocorrelation function of one of two beams generated in type-II parametric down-conversion by means of temporal magnification of this beam, bringing its correlation time from the picosecond to the nanosecond scale, which can be resolved by modern photodetectors. We show that such a measurement enables one to infer directly the degree of global coherence of that beam, which is linked by a simple relation to the number of modes characterizing the entanglement between the two generated beams. We illustrate the proposed method by an example of photon pairs generated in a periodically poled potassium titanyl phosphate (KTP) crystal with a symmetric group velocity matching for various durations of the pump pulse, resulting in different numbers of modes. Our theoretical model also shows that the magnified double-heralded autocorrelation function of one beam exhibits a local maximum around zero delay time, corresponding to photon bunching at a short time scale. (10.1103/7ckm-tm3r)
  DOI : 10.1103/7ckm-tm3r
• SpectreShield: Design and Analysis of Spectre Countermeasures on RISC-V Using gem5
  
  • Khan Mahreen
  • Mushtaq Maria
  • Pacalet Renaud
  • Apvrille Ludovic
  , 2025. <div><p>Speculative execution attacks like Spectre exploit microarchitectural side effects to leak sensitive data during transient execution. While various software and hardware countermeasures have been proposed for x86 and ARM architectures, their effectiveness and microarchitectural impact remain underexplored on RISC-V platforms. To study such attacks and evaluate these countermeasures, simulation tools like the gem5 simulator provide detailed insights into microarchitectural state changes during speculation. In this paper, we present the first comprehensive evaluation of Spectre-v1 countermeasures on the RISC-V architecture using the gem5 full-system simulator. We implement and assess four Spectre-v1 mitigations: index masking (CM1), randomized offset (CM2), fence-based serialization (CM3), and bitwise selection (CM4). Our experiments reveal that, in the absence of mitigations, Spectre-v1 enables 100% secret key recovery. In contrast, all proposed countermeasures reduce the recovery rate to below 1%, with branch mispredictions decreasing by 41.7%-46.3%. The paper analyzes the securityperformance trade-offs of each approach. Beyond demonstrating their effectiveness, we quantify their microarchitectural impact, measuring reductions in squashed instructions, DRAM latency variability, and return address stack mispredictions. This paper provides a practical framework for evaluating transient execution defenses and advances secure-by-design RISC-V processors.</p></div>
• Satellite Image Time-Series Data Augmentation Using an Attention Mechanism Variational Recurrent Autoencoder
  
  • Chaabane Ferdaous
  • Tupin Florence
  , 2025. Data scarcity presents a significant challenge in satellite image analysis, particularly for developing robust models in remote sensing applications. High-quality and abundant data are essential for accurate predictions; however, acquiring Satellite Image Time-Series (SITS) data is often constrained by factors such as limited temporal coverage and the high cost of Very High Resolution (VHR) acquisitions. To address this issue, we propose a novel Attention-based Variational Recurrent Autoencoder (AVRAE) designed for generating synthetic satellite image time-series data. This method extends the evidence lower bound (ELBO) of variational inference to incorporate the temporal dependencies essential for satellite data. A recurrent neural network-based autoencoder framework is employed, integrated with an attention mechanism to effectively capture both short-and long-term temporal relationships. The AVRAE framework synthesizes realistic and statistically representative satellite time-series data, enabling enhanced analysis for remote sensing applications. Evaluations using real-world satellite datasets demonstrate that AVRAE produces coherent and statistically valid synthetic data, thereby improving VHR SITS data quality for deep learning-based remote sensing applications.
• Bayesian Stream Tuner: Dynamic Hyperparameter Optimization for Real-Time Data Streams
  
  • Verma Nilesh
  • Bifet Albert
  • Pfahringer Bernhard
  • Bahri Maroua
  , 2025, 2, pp.2871-2882. Hyperparameter optimization is crucial for maximizing machine learning model performance, yet most existing algorithms are designed for batch or offline scenarios and assume static data distributions. Such assumptions fall short in data stream settings, where models must adapt to evolving inputs in real time. To address these limitations, we propose the Bayesian Stream Tuner (BST), a novel framework for online hyperparameter optimization in nonstationary data streams. BST maintains a dynamic set of candidate hyperparameter configurations and periodically refines them using an incremental Bayesian model, which estimates configuration performance based on recent data statistics and hyperparameter values. This systematic exploration and refinement strategy allows BST to detect and respond to concept drift by resetting its adaptation mechanisms whenever necessary, ensuring strong performance under changing distributions. Our theoretical analysis establishes sublinear regret bounds for BST in dynamic environments, and extensive experiments on classification and regression tasks demonstrate that BST consistently outperforms state-of-the-art online hyperparameter optimization methods in both predictive accuracy and adaptability, making it a powerful solution for real-time hyperparameter tuning in evolving data streams. (10.1145/3711896.3736852)
  DOI : 10.1145/3711896.3736852
• An Information Theoretic Proof of the Chernoff-Hoeffding Inequality
  
  • Rioul Olivier
  • Solé Patrick
  Information Processing Letters, Elsevier, 2025, 190, pp.106582. The Chernoff bound is a well-known upper bound on the tail of binomial distributions of parameter 1/2 involving the binary entropy function. Hoeffding's inequality (or the Chernoff-Hoeffding inequality) is a generalization for binomial distributions of parameter 1 -1/q, involving the q-ary entropy function (with q ≥ 2), which can be written in terms of the Kullback-Leibler divergence and is related to the bound in Fano's inequality. We give an information theoretic proof of that bound, and sketch some applications to channel and source coding. We also derive a refined bound which is always sharper. (10.1016/j.ipl.2025.106582)
  DOI : 10.1016/j.ipl.2025.106582
• Long run convergence of discrete-time interacting particle systems of the McKean-Vlasov type
  
  • Bianchi Pascal
  • Hachem Walid
  • Priser Victor
  Stochastic Processes and their Applications, Elsevier, 2025. We consider a discrete-time system of n coupled random vectors, a.k.a. interacting particles. The dynamics involve a vanishing step size, some random centered perturbations, and a mean vector field which induces the coupling between the particles. We study the doubly asymptotic regime where both the number of iterations and the number n of particles tend to infinity, without any constraint on the relative rates of convergence of these two parameters. We establish that the empirical measure of the interpolated trajectories of the particles converges in probability, in an ergodic sense, to the set of recurrent Mc-Kean-Vlasov distributions. A first application example is the granular media equation, where the particles are shown to converge to a critical point of the Helmholtz energy. A second example is the convergence of stochastic gradient descent to the global minimizer of the risk, in a wide two-layer neural networks using random features.
• On the spectral decomposition of the complex Robin Laplacian
  
  • Badeau Roland
  Journal of the Acoustical Society of America, Acoustical Society of America, 2025, 158 (1), pp.838-848. The mathematical properties of the Laplacian on a bounded domain are well-known when the boundary condition is of the first type (Dirichlet) or second type (Neumann). In both cases, this operator is self-adjoint and, therefore, diagonalizable, its spectrum is discrete, and the set of eigenfunctions can be chosen to form an orthonormal basis of the Hilbert space of square-integrable functions on the domain. However, in the case of the third type (Robin) boundary condition, the same is true only when the parameter is real-valued. On the contrary, when this parameter is complex-valued, the Laplacian may not even be diagonalizable. In this paper, the spectral decomposition of the complex Robin Laplacian is investigated in the most general case possible, and a formula that decomposes any square-integrable function on the set of its (generalized) eigenfunctions is provided. This result is applied to the Green's function of the Helmholtz equation, whose existence, unicity, and closed-form expression are established in this general setting, and the statistical wave field theory, which provides the statistical laws of waves propagating in a bounded domain. (10.1121/10.0037233)
  DOI : 10.1121/10.0037233
• Melody-Lyrics Matching with Contrastive Alignment Loss
  
  • Wang Changhong
  • Olvera Michel
  • Richard Gaël
  , 2025. The connection between music and lyrics is far beyond semantic bonds. Conceptual pairs in the two modalities such as rhythm and rhyme, note duration and syllabic stress, and structure correspondence, raise a compelling yet seldom-explored direction in the field of music information retrieval. In this paper, we present melody-lyrics matching (MLM), a new task which retrieves potential lyrics for a given symbolic melody from text sources. Rather than generating lyrics from scratch, MLM essentially exploits the relationships between melody and lyrics. We propose a self-supervised representation learning framework with contrastive alignment loss for melody and lyrics. This has the potential to leverage the abundance of existing songs with paired melody and lyrics. No alignment annotations are required. Additionally, we introduce sylphone, a novel representation for lyrics at syllable-level activated by phoneme identity and vowel stress. We demonstrate that our method can match melody with coherent and singable lyrics with empirical results and intuitive examples. We open source code and provide matching examples on the companion webpage: https://github.com/changhongw/mlm.
• Benchmarking the Benchmarks: Reproducing Climate-Related NLP Tasks
  
  • Calamai Tom
  • Balalau Oana
  • Suchanek Fabian M
  , 2025. Significant efforts have been made in the NLP community to facilitate the automatic analysis of climate-related corpora by tasks such as climate-related topic detection, climate risk classification, question answering over climate topics, and many more. In this work, we perform a reproducibility study on 8 tasks and 29 datasets, testing 6 models. We find that many tasks rely heavily on surface-level keyword patterns rather than deeper semantic or contextual understanding. Moreover, we find that 96% of the datasets contain annotation issues, with 16.6% of the sampled wrong predictions of a zero-shot classifier being actually clear annotation mistakes, and 38.8% being ambiguous examples. These results call into question the reliability of current benchmarks to meaningfully compare models and highlight the need for improved annotation practices. We conclude by outlining actionable recommendations to enhance dataset quality and evaluation robustness.
• Graphically Speaking: Unmasking Abuse in Social Media with Conversation Insights
  
  • Nouri Célia
  • Cointet Jean-Philippe
  • Clavel Chloé
  , 2025. Detecting abusive language in social media conversations poses significant challenges, as identifying abusiveness often depends on the conversational context, characterized by the content and topology of preceding comments. Traditional Abusive Language Detection (ALD) models often overlook this context, which can lead to unreliable performance metrics. Recent Natural Language Processing (NLP) approaches that incorporate conversational context often rely on limited or overly simplified representations of this context, leading to inconsistent and sometimes inconclusive results. In this paper, we propose a novel approach that utilizes graph neural networks (GNNs) to model social media conversations as graphs, where nodes represent comments, and edges capture reply structures. We systematically investigate various graph representations and context windows to identify the optimal configurations for ALD. Our GNN model outperforms both context-agnostic baselines and linear context-aware methods, achieving significant improvements in F1 scores. These findings demonstrate the critical role of structured conversational context and establish GNNs as a robust framework for advancing context-aware ALD. Our code is available at https://github.com/celia-nouri/ConversationALD/.
• StreamMLOps: Online Learning in Practice from Big Data Streams &amp; Real-Time Applications
  
  • Barry Mariam
  • Montiel Jacob
  • Bifet Albert
  • Manchev Nikolay
  • Wadkar Sameer
  • Halford Max
  • Chiky Raja
  • El Jaouhari Saad
  • Shakman Katherine B
  • Al Fehaily Joudi
  • Le Deit Fabrice
  • Tran Vinh-Thuy
  • Guerizec Eric
  , 2025. <div><p>Learning and serving from evolving streaming data to real-time inference in production is a challenging problem. Traditionally, data is partitioned and processed in batches to train machine learning models. In dynamic environments, models' performance drops over time (model degradation), requiring new models to be trained and deployed in their place. This paper deals with the MLOps aspects of deploying online and continual learning models addressing the requirements in the production of real-time applications. We have demonstrated that Online Learning methods can be scaled horizontally in production to meet the high-velocity streaming feature pipeline. The design is based on open platforms and the paper demonstrates an MLOps strategy to execute Online Learning and Predictions, perform Online Learning on a stream and deploy an online learning model version without stream interruption. The approach is suitable for highly regulated industries like banking which also have high throughput requirements. Experiments on high-dimensional and feature-evolving data streams (Malicious URL detection) demonstrate the effectiveness and efficiency of online learning models in terms of time, space and F1-score. Finally, we provide some best practices for using architectural design to deploy these dynamic models on a stream and perform Online Learning and deploy them without stopping the streaming pipeline using open-source technology such as Kafka, Flink, MLflow and river.</p></div> (10.1109/ICDE55515.2023.00272)
  DOI : 10.1109/ICDE55515.2023.00272
• Survey on forecasting for electric vehicle charging-power demand
  
  • Yang Wen
  • Laurenty Ignacio
  • Fontaine Mathieu
  • d'Alché-Buc Florence
  , 2025.
• Investigating Raman backscattering decay and the perspective of time-multiplexed quantum communications
  
  • Verdier Pierre-Enguerrand
  • Alléaume Romain
  • Rivera Thomas
  Optics Express, Optical Society of America - OSA Publishing, 2025, 33 (15), pp.31029-31041. We have studied the temporal dynamics of Raman scattering caused by classical power in optical fiber and its impact on counter-propagating quantum signals. We investigated, on the entire telecom bands, the duration during which the quantum channel cannot be used in a time-division multiplexing context. Thereby, we estimated performance in terms of secure key rates within the framework of time-division multiplexing. By applying our model to the discrete variable quantum key distribution (DV-QKD) protocol BB84 in different optical communication contexts, we demonstrate the feasibility of counter-propagating time-multiplexing classical and quantum communications. Our results highlight a better preservation of the maximum communication distance for quantum channels compared to other multiplexing schemes. (10.1364/OE.561961)
  DOI : 10.1364/OE.561961
• Don’t Forget Your Inverse DDIM for Image Editing
  
  • Gomez-Trenado Guillermo
  • Mesejo Pablo
  • Cordón Oscar
  • Lathuilière Stéphane
  IEEE Computational Intelligence Magazine, Institute of Electrical and Electronics Engineers, 2025, 20 (3), pp.10-18. The field of text-to-image generation has undergone significant advancements with the introduction of diffusion models. Nevertheless, the challenge of editing real images persists, as most methods are either computationally intensive or produce poor reconstructions. This paper introduces SAGE (Self-Attention Guidance for image Editing) - a novel technique leveraging pre-trained diffusion models for image editing. SAGE builds upon the DDIM algorithm and incorporates a novel guidance mechanism utilizing the self-attention layers of the diffusion U-Net. This mechanism computes a reconstruction objective based on attention maps generated during the inverse DDIM process, enabling efficient reconstruction of unedited regions without the need to precisely reconstruct the entire input image. Thus, SAGE directly addresses the key challenges in image editing. The superiority of SAGE over other methods is demonstrated through quantitative and qualitative evaluations and confirmed by a statistically validated comprehensive user study, in which all 47 surveyed users preferred SAGE over competing methods. Additionally, SAGE ranks as the top-performing method in seven out of 10 quantitative analyses and secures second and third places in the remaining three. (10.1109/MCI.2025.3563859)
  DOI : 10.1109/MCI.2025.3563859
• Routing Quantum Control of Causal Order
  
  • Grothus Maarten
  • Abbott Alastair A.
  • Vanrietvelde Augustin
  • Branciard Cyril
  , 2025. In recent years, various frameworks have been proposed for the study of quantum processes with indefinite causal order. In particular, quantum circuits with quantum control of causal order (QC-QCs) form a broad class of physical supermaps obtained from a bottom-up construction and are believed to represent all quantum processes physically realisable in a fixed spacetime. Complementarily, the formalism of routed quantum circuits introduces quantum operations constrained by "routes" to represent processes in terms of a more fine-grained routed circuit decomposition. This decomposition, formalised using a so-called routed graph, represents the information flow within the respective process. However, the existence of routed circuit decompositions has only been established for a small set of processes so far, including both certain specific QC-QCs and more exotic processes as examples. In this work, we remedy this fact by connecting these two frameworks. We prove that for any given $N$, one can use a single routed graph to systematically obtain a routed circuit decomposition for any QC-QC with $N$ parties. We detail this construction explicitly and contrast it with other routed circuit decompositions of QC-QCs, which we obtain from alternative routed graphs. We conclude by pointing out how this connection can be useful to tackle various open problems in the field of indefinite causal order, particularly establishing circuit representations of subclasses of QC-QCs.
• Assessing the Vulnerabilities of RISC-V using the gem5 Simulator
  
  • Khan Mahreen
  • Mushtaq Maria
  • Pacalet Renaud
  • Apvrille Ludovic
  , 2025. <div><p>Microarchitectural timing side-channel attacks exploit variations in execution times caused by the underlying hardware to extract sensitive information. These attacks leverage architectural features like caches, branch predictors, and speculative execution.</p></div>
• Winner-takes-all for Multivariate Probabilistic Time Series Forecasting
  
  • Cortés Adrien
  • Rehm Rémi
  • Letzelter Victor
  , 2025. We introduce TimeMCL, a method leveraging the Multiple Choice Learning (MCL) paradigm to forecast multiple plausible time series futures. Our approach employs a neural network with multiple heads and utilizes the Winner-Takes-All (WTA) loss to promote diversity among predictions. MCL has recently gained attention due to its simplicity and ability to address ill-posed and ambiguous tasks. We propose an adaptation of this framework for time-series forecasting, presenting it as an efficient method to predict diverse futures, which we relate to its implicit quantization objective. We provide insights into our approach using synthetic data and evaluate it on real-world time series, demonstrating its promising performance at a light computational cost.
• Two-Stage MIMO Equalization for Long Haul Coupled Multi-Core Fiber Systems
  
  • Darweesh Jamal
  • Abouseif Akram
  • Rekaya Ben Othman Ghaya
  • Jaouën Yves
  • Klaimi Rami
  , 2025.
• Beyond Low-rank Decomposition: A Shortcut Approach for Efficient On-Device Learning
  
  • Nguyen Le-Trung
  • Quélennec Aël
  • Nguyen van Tam
  • Tartaglione Enzo
  , 2025, Forty-second International Conference on Machine Learning. On-device learning has emerged as a promising direction for AI development, particularly because of its potential to reduce latency issues and mitigate privacy risks associated with device-server communication, while improving energy efficiency. Despite these advantages, significant memory and computational constraints still represent major challenges for its deployment. Drawing on previous studies on low-rank decomposition methods that address activation memory bottlenecks in backpropagation, we propose a novel shortcut approach as an alternative. Our analysis and experiments demonstrate that our method can reduce activation memory usage, even up to 120.09× compared to vanilla training, while also reducing overall training FLOPs up to 1.86× when evaluated on traditional benchmarks.
• To Each Metric Its Decoding: Post-Hoc Optimal Decision Rules of Probabilistic Hierarchical Classifiers
  
  • Plaud Roman
  • Perez-Lebel Alexandre
  • Labeau Matthieu
  • Saillenfest Antoine
  • Bonald Thomas
  , 2025. Hierarchical classification offers an approach to incorporate the concept of mistake severity by leveraging a structured, labeled hierarchy. However, decoding in such settings frequently relies on heuristic decision rules, which may not align with task-specific evaluation metrics. In this work, we propose a framework for the optimal decoding of an output probability distribution with respect to a target metric. We derive optimal decision rules for increasingly complex prediction settings, providing universal algorithms when candidates are limited to the set of nodes. In the most general case of predicting a subset of nodes, we focus on rules dedicated to the hierarchical hF β scores, tailored to hierarchical settings. To demonstrate the practical utility of our approach, we conduct extensive empirical evaluations, showcasing the superiority of our proposed optimal strategies, particularly in underdetermined scenarios. These results highlight the potential of our methods to enhance the performance and reliability of hierarchical classifiers in real-world applications. The code is available at https://github.com/RomanPlaud/ hierarchical decision rules Sketch of the proof. Key elements of the decoding strategy are displayed in Algorithm 1. We give here some general insights on how the algorithm is derived.
• Assessing the Vulnerabilities of RISC-V using the gem5 Simulator (Access-Retired)
  
  • Khan Mahreen
  • Mushtaq Maria
  • Pacalet Renaud
  • Apvrille Ludovic
  , 2025. <div><p>Emerging RISC-V processors require rigorous security evaluation to address microarchitectural vulnerabilities inherent in their rapidly evolving ecosystem. A recent paper [Gea23] implemented both known and novel side-channel attacks targeting commercial RISC-V CPUs (U74 and C906). While this hardware-based research confirmed vulnerabilities, it could not provide detailed insights into attack dynamics. We bridge this gap using the gem5 simulation framework to systematically analyze side-channel attacks on RISC-V architectures. Our paper focuses on the accessretired attack, which exploits the unprivileged rdinstret instruction to infer protected filesystem data. By tracking retired instruction counts, attackers detect microarchitectural state differences caused by directory access checks. We utilize the gem5 simulator in full-system (FS) mode to capture kernel-level behaviors, allowing us to analyze critical performance metrics including instruction retirement, cache performance, and branch prediction statistics. This detailed simulationbased analysis is essential for understanding the behavior of the attack and for developing effective countermeasures. Advancing RISC-V security research with simulation tools like gem5 is thus a promising direction for mitigating future side-channel vulnerabilities.</p></div>
• DRST: a Non-Intrusive Framework for Performance Analysis in Softwarized Networks
  
  • Liu Qiong
  • Lin Jianke
  • Zhang Tianzhu
  • Linguaglossa Leonardo
  , 2025. <div><p>The last decade has witnessed the proliferation of network function virtualization (NFV) in the telco industry, thanks to its unparalleled flexibility, scalability, and cost-effectiveness. However, as the NFV infrastructure is shared by virtual network functions (VNFs), sporadic resource contentions are inevitable. Such contention makes it extremely challenging to guarantee the performance of the provisioned network services, especially in high-speed regimes (e.g., Gigabit Ethernet). Existing solutions typically rely on direct traffic analysis (e.g., packet-or flow-level measurements) to detect performance degradation and identify bottlenecks, which is not always applicable due to significant integration overhead and systemlevel constraints. This paper complements existing solutions with a lightweight, non-intrusive framework for online performance inference that easily adapts to drift (i.e., a change over time of the actual state of our system). Instead of direct data-plane collection, we reuse hardware features in the underlying NFV infrastructure, introducing negligible interference in the data-plane. Our Drift-Resilient and Self-Tuning (DRST) framework can be integrated into existing NFV systems with minimal engineering effort and operates without the need for predefined traffic models or VNF-specific customization. DRST is deployed via a lightweight MLOps pipeline that automates the adaptation under runtime drift. We show how DRST can deliver accurate performance inference or diagnose run-time bottleneck diagnose, as demonstrated through comprehensive evaluation across diverse NFV scenarios.</p></div>
• SimHawNet: a modified Hawkes process for temporal network simulation
  
  • Perez Mathilde
  • Romero Raphaël
  • Kang Bo
  • de Bie Tijl
  • Lijffijt Jefrey
  • Laclau Charlotte
  Data Mining and Knowledge Discovery, Springer, 2025, 39 (5), pp.48 (1-28 pages). Abstract Temporal networks allow representing connections between objects while incorporating the temporal dimension. While static network models can capture unchanging topological regularities, they often fail to model the effects associated with the causal generative process of the network that occurs in time. Hence, exploiting the temporal aspect of networks has been the focus of many recent studies. In this context, we propose a new framework for generative models of continuous-time temporal networks. We assume that the activation of the edges in a temporal network is driven by a specified temporal point process. This approach allows to directly model the waiting time between events while incorporating time-varying history-based features as covariates in the predictions. Coupled with a thinning algorithm designed for the simulation of point processes, SimHawNet enables simulation of the evolution of temporal networks in continuous time. Finally, we introduce a comprehensive evaluation framework to assess the performance of such an approach, in which we demonstrate that SimHawNet successfully simulates the evolution of networks with very different generative processes and achieves performance comparable to the state of the art, while being significantly faster. (10.1007/s10618-025-01119-1)
  DOI : 10.1007/s10618-025-01119-1
• Défi du GDR GPL ADaptation DYnamique et ConTinue ADDYCT
  
  • Laval Jannik
  • Philippe Jolan
  • Cariou Eric
  • Ameur-Boulifa Rabéa
  • Guérin Sylvain
  • Kouchnarenko Olga
  • Guermouche Nawal
  , 2025. L'adaptabilité est un enjeu majeur des systèmes complexes dans des environnements dynamiques. Ces environnements regroupent les architectures distribuées composées de systèmes "component-based" et les infrastructures déployées sur des plateformes hétérogènes à différentes échelles : Cloud, Fog, Edge, ou IoT. Tous ces systèmes doivent être capables d'ajuster leur configuration de manière autonome pour répondre à des évènements exogènes et/ou endogènes. Les systèmes logiciels doivent être considérés dès leur conception comme des systèmes durables en termes de temporalité (Système temps long, Cycle de vie, Couplage), de scalabilité (Granularité, Interfaces, Gestion massive de données) et d'hétérogénéité (Intégration, Interopérabilité). Ces challenges sont d'autant plus importants lorsque la taille du système est grande et couplée avec des artefacts matériels (IoT, CPS, Jumeaux numériques, Cloud...) L'objectif de ce défi est de modéliser, analyser et d'implémenter des moyens et des politiques d'adaptation pour des systèmes logiciels complexes (distribués, componentisés etc.). L'approche adoptée repose sur les boucles de contrôle MAPE-K, un modèle d'auto-adaptation autonome capable de s'ajuster dynamiquement à un environnement permettant de répondre à la nécessité d'intégrer les données collectées et leurs modèles de traitement, le système opérant et sa connexion avec son environnement, l'évolution dynamique et la nécessité de maintenir une représentation fidèle du comportement attendu.
• Deep Learning for Embedded Cybersecurity
  
  • Varillon Arnaud
  , 2025. Public-key cryptography is one of the core pillars of cybersecurity, in particular thanks to the authentication schemes it enables. It is embedded in many ubiquitous objects, such as hardware wallets. Side-channel attacks are one of the major threats to such devices. In particular, the advances in machine learning, and more specifically deep learning, which have marked the last ten years, seem likely to make such attacks remarkably effective. In such a hostile environment, assessing the true security level of devices intended for cryptographic use is of utmost importance: indeed, it has now even become vital to the smooth running of information systems.In this thesis, we have appraised the security, of implementations, reputed to be the more secure ones in the face of such attacks (“power” and “EM” channels), which manipulate the secret key bit by bit. Numerous contributions, which mainly use deep learning, have been published on this subject. Unfortunately, none of them provides any guarantee regarding the robustness of the device under study in the face of such attacks: each time, the method being described does not allow one to state with certainty that it is not possible to find a more powerful attack than that being presented. The security level of the latter devices is therefore potentially underestimated. More specifically, fundamental aspects of the classification task associated with any attack, such as the shape of its decision boundary or the optimality - in terms of attack performance - of the features derived from the sampling of the side-channel under consideration, are never addressed. Therefore, we have sought to find methodologies that are as close to optimality as possible given the conditions imposed by the exercise (for example, the possibility of configuring the key used by the target for in-depth analyses).Initially, assuming that an attacker can control the key that is parameterized in the target, only vertical attacks are considered. In this context, the optimal effectiveness of the joint use of NICV, for feature selection, and the perceptron, for classification, is highlighted from a theoretical point of view. In particular, the security of a cryptographic library hitherto considered robust (libecc) is called into question. Secondly, assuming that an attacker cannot set the key as he wishes in the target, yet still has a functionally perfect clone, another procedure is proposed for carrying out the security evaluation, this time using horizontal (collision-based) attacks based on an unsupervised learning technique which, because it requires (by definition) minimal training at most, is better suited to such a scenario. Compared to the state of the art, the approach followed is closer to optimality without however achieving it, but avenues are suggested to get there in the near future. Last, to validate these findings, experimental verifications have been carried out on a board (STM32F407) which features a Cortex-M4 processor that can be found in many hardware wallets (e.g. Trezor Model T).