HAL-CentraleSupelec
Not a member yet
77624 research outputs found
Sort by
Modeling turbulent flame enhancement by Nanosecond Repetitively Pulsed discharges using a low-order model
No full textInternational audiencePlasma-Assisted Combustion (PAC), using Nanosecond Repetitively Pulsed (NRP) discharges, is a promising technique to stabilize lean premixed flames, which are prone to instabilities and extinction. PAC has been successfully demonstrated in various academic and semi-industrial configurations. It has been proven to be effective in preventing instabilities, improving combustion efficiency, and extending the lean blowout (LBO) limit. To transfer PAC technology toward higher Technology Readiness Levels (TRLs), numerical simulations are required by engineers for combustor design and optimization. Among the strategies available, several multi-D Direct Numerical Simulations (DNS) and Large Eddy Simulations (LES) of plasma-assisted turbulent combustion have been conducted in the literature by combining phenomenological NRP discharge models with detailed or semi-detailed combustion mechanisms. While these simulations achieve good accuracy, their computational cost remains high because of the large variety of spatial and temporal scales involved in plasma, combustion, and turbulence interactions. The present work aims to develop a low-order model of flame stabilization by NRP discharges, as alternative to full 3-D CFD simulations, for the design and optimization of PAC systems at a low-CPU cost. PAC is modeled by a series of three connected canonical flame elements. First, the volume of gas affected by the plasma discharges is modeled by a Perfectly-Stirred Reactor (PSR), which employs Castela's phenomenological NRP discharge model. Next, a Plug Flow Reactor (PFR) is employed to track the combustion reactions within the recirculation zone. Finally, a 1-D strain-imposed Premixed Counterflow Flame (PMX-CF) models the impact of these gas fluxes on the flame structure. The reduced-order model is validated against 3-D LES of the Mini-PAC configuration, and a parametric study is performed on some key modeling parameters, namely the dilution and the strain rate
PPIR: A Privacy-Preserving and Intrusion-Resilient decentralized gossig-based trajectory prediction framework for VANETs
No full textInternational audienceSecurity breaches in Vehicular Ad-hoc Networks (VANETs) pose significant threats by compromising sensitive data and endangering the physical safety of road users and their property. This issue becomes particularly critical when the vehicle trajectory prediction process is compromised. Thus, ensuring the security of autonomous vehicles has become a significant challenge. In this context, we introduce PPIR (Privacy-Preserving and Intrusion-Resilient decentralized gossig-based trajectory prediction framework for VANETs) , a lightweight and robust framework to detect and mitigate malicious behavior, reinforced by privacy-preserving mechanisms for distributed and collaborative trajectory prediction in VANETs. The proposed approach integrates a data-driven misbehavior detection and mitigation algorithm, coupled with a unified elliptic curve-based encryption and signature process. This algorithm relies on innovation errors computed through an extended Kalman filter to assess the consistency and plausibility of the parameters received by each vehicle from peers. The system is fully decentralized and operates using a gossip-based communication protocol, making it wellsuited for highly dynamic vehicular environments. Experimental results demonstrate that PPIR achieves superior performance compared to existing data-centric misbehavior detection systems, while remaining efficient under the computational constraints inherent to vehicular networks.</div
Long-term useful data rate optimization of batteryless devices powered by intermittent energy sources
No full textInternational audienceThis paper explores the optimization of the longtermuseful data rate of a communicating device supplied by intermittentenergy sources. Based on experimental measurementsof the energy consumed for communication, an energy model isestablished for a simple node transmitting data towards an accesspoint using a communication protocol. First, the long-term usefuldata rate is optimized for a fixed energy storage, demonstratingthat the optimal rate depends only on the average power providedby the source. Second, a time constraint is introduced for whichinformation must be sent at least in a given time slot. The solutionof this time constrained problem is determined by the statisticalbehavior of the energy source. The numerical results are validatedthrough Monte Carlo simulations. Finally, design guidelines areprovided to develop efficient communicating devices powered byintermittent energy sources
Evaluating the Realism of Cyber-Physical Honeynets Against Advanced Attackers
No full textCyber-physical honeynets are increasingly deployed to study adversarial behavior in operational technology (OT) and industrial control systems (ICS), yet their effectiveness depends on their perceived realism. This work presents a multi-stage research program systematically characterizing, measuring, and quantifying the realism of cyber-physical honeynets against advanced attackers. First, we conduct two systematic literature reviews: one on cyber-physical honeynets, producing an updated taxonomy and a reference architecture, and another on anti-honeypot techniques, revealing a critical gap between academic detection methods and real-world adversarial practices. We then empirically investigate attacker behavior through a large-scale Capture-the-Flag (CTF) experiment, analyzing 8,544 shell commands to identify real-world anti-honeypot strategies and behavioral indicators of perceived authenticity. Building on these insights, we propose a novel evaluation methodology using real threat actors and ICS-targeting malware to derive quantitative metrics of honeypot realism. Finally, we outline the design of an automated pentesting framework that operationalizes validated detection techniques to compute a reproducible Realism Score for heterogeneous honeypot deployments
Fault diagnosis through system-level condition-monitoring and digital-twin-supported deep learning
No full textDiagnosing component-level faults in complex systems is a challenging task, particularly when only systemlevel condition-monitoring data are available due to difficulties in deploying sensors at the component level.To address this issue, we propose a digital twin-supported deep learning framework for diagnosing componentlevel failures using system-level condition-monitoring data. The framework operates in two phases: an offline phase, where a Digital Failure Twin (DFT) model simulates failure modes and generates synthetic training data, and an online phase, where a sim-to-real error correction mechanism aligns simulation outputs with real-world system behavior using an error simulator and hyper-parameter tuning. This alignment ensures the diagnostic model effectively bridges the gap between simulated and real-world conditions. The proposed framework is evaluated on a real-world robot system where only the movement trajectory of the end-effector is used as condition-monitoring data to diagnose the failure modes of the four motors of the robot. An open-source DFT model for robotic systems was developed to generate synthetic failure data for training the diagnosis models, while real-world test data were collected to assess the model's performance. Experimental results demonstrate the effectiveness of the developed framework: it improves fault diagnosis accuracy on real datasets by up to 403.76% as compared to traditional deep learning models without using the synthetic data from the DFT, especially in scenarios with limited real data. Furthermore, with only 10 real operational samples, the developed sim-to-real correction method improves accuracy on the real test data by 15.51%.</div
Case-Based Identification of Anti-Coagulation Therapy for Pregnant Women Using Approximation and Interpolation of Sequences (extended version)
No full textIn a current medical study about anti-coagulation therapy for pregnant women, an issue of missing data has been raised. This issue is modeled as a problem-solving task where a problem is a description of the known data of a patient and the solution is her missing data. A case-based approach is used for inferring missing data, where a case is a complete data. The approaches of approximation (using a similarity relation) and interpolation (using a fuzzy betweenness relation) are considered for implementing such a CBR system. A technical difficulty is that the problems in this application are represented using temporal sequences, and the issue of implementing similarity and betweenness on such sequences had to be addressed. The proposed approach is based on similarity paths, i.e. paths in the problem space. The evaluation had to be made on generated data but has shown the efficiency of these approaches.</div
Code-switching as a Bias Indicator in LLMs: "The consequences are not the same para nosotros"
No full textInternational audienceCode-switching is a widespread linguistic practice among bilingual speakers. While recent studies have addressed the impact of code-switching on downstream task performance, the potential biases and harms that language models may cause when prompted with code-switching have yet to be investigated. The objective of this study is to investigate whether code-switching constitutes an implicit indicator of ethnicity that can be leveraged to unveil covert racist or xenophobic bias in language models. The present paper introduces a methodology to compare generated texts that were prompted with code-switching vs. with monolingual inputs. It is applied on both Hinglish and Spanglish, two popular forms of code-switching that are omnipresent in Indian and Hispanic communities. With a decision tree approach, we tackle various types of semantic differences through the use of semantic resources, stereotypes lists, POS-tagging and sentiment classifiers. Over 84k text pairs are generated with 3 popular large language models. Overall, around 50% of generated text pairs are not semantically equivalent, and 25% of the time, there is a potential for harm against the Indian or Hispanic community. The different possible harms are further discussed, relying on sociological studies to argue that bias and harms against socially discriminated communities have greater consequences
Electrode shading lithography: A new method to avoid side dissolution of adhesion layer during electrodeposition of platinum black
No full textInternational audienc
MOQ-n-Roll: Live Sports Watch Parties
No full textInternational audienceLive sports watch parties must deliver broadcast content reliably while enabling low-latency group interaction. We address these conflicting requirements using MOQ Transport (MOQT) through a dual-pipe architecture. The context pipe carries the main program over reliable QUIC streams, enforcing a spoiler prevention model that keeps participants synchronized. The excitement pipe carries participant audio and video, selecting transport primitives by modality: unreliable datagrams minimize delay for independent audio frames, while multiplexed (reliable) streams with early discard preserve decoding order for video without accumulating latency
A measurement-based calibration approach for highly scalable timing and energy modeling of EdgeAI multi-core systems
No full textInternational audienceDeploying Artificial Neural Networks (ANNs) on embedded multi-core platforms requires precise models for estimating and optimizing timing and energy, which is crucial for enabling novel Artificial Intelligence (AI) applications. However, predicting non-functional properties (timing, power) is challenging due to degrees of parallelism in ANNs and complex effects in execution platforms (e.g. contentions at shared resources, dynamic power management). This article presents an Electronic System-Level (ESL) timing and energy modeling flow and the associated calibration methodology for optimizing ANN deployment on multi-core platforms. The proposed flow leverages SystemC simulation to offer both speed and accuracy while ensuring high scalability in many dimensions, such as platform resources modeling. Analytical models are used for ANN layer computation and communication delays as well as power consumption and energy cost. We propose a measurement-based calibration approach to these models which enables high prediction accuracy while guaranteeing high re-usability. The calibrated models can be used across different settings without the need to re-perform a calibration phase. We validate our flow against real measurements of ANN implementations on a prototype multi-core platform. Results demonstrate over 97% accuracy in timing and 93% in energy for 54 mappings of different ANNs tested with and without the use of power management on the platform, with an evaluation time under 2s per mapping. Furthermore, we illustrate that our flow is suitable for Design Space Exploration (DSE), allowing up to 24% improvement in inference time and 16% in energy compared to baseline implementation