Université de technologie de Troyes open archive
Not a member yet
10722 research outputs found
Sort by
A multi-agent federated reinforcement learning-based optimization of quality of service in various LoRa network slices
No full textInternational audienceThe innovations heralded through the implementation of next-generation 5G (Fifth-Generation) networks provide an opportunity for the efficient coexistence of heterogeneous services distributed by a single physical virtualized infrastructure. Indeed, thanks to the possibility of virtualization of network functions implemented in 5G networks, the management of physical resources will be able to become more flexible and users will be able to benefit from a service customization to satisfy their demand in terms of energy efficiency, throughput and communication reliability. However, these advances are not without constraints. The management of physical and virtual resources will become more complex given the number of connected objects which will increase, generating a large volume of data to manage. This therefore requires the implementation of much more intelligent systems in the network controllers in order to guarantee the QoS (Quality of Service) of the communications. Thus, an important axis of research is now oriented towards artificial intelligence techniques, more precisely reinforcement learning to overcome this problem. Driven by this context, we are directing our research towards improving the QoS offered to users of connected objects by proposing an optimization model based on network slicing and federated reinforcement learning in order to minimize energy consumption, maximize user throughputs and reduce latency during communications between LoRa (Long Range) devices. The results obtained by the simulations carried out in a realistic framework clearly demonstrate that our proposal optimizes the traffic in each network slice and also for the individual user
IoV security and privacy survey: issues, countermeasures, and challenges
No full textInternational audienceAs a growing up-and-coming branch of the Internet of Things and traditional vehicular ad hoc networks, the Internet of Vehicles (IoV) is intended to perform as a core information carrying and processing platform for Intelligent Transport Systems. However, owing to its dynamic topological structures, large network scale, and mobile limitation, IoV systems still struggle with many unresolved challenges, especially those concerning security and privacy. Therefore, researchers consider security a significant concern due to the variety of vulnerabilities. Recently, existing security researchers have worked extensively to guarantee IoV security. However, multiple challenges in terms of security and privacy are generated by various types of attacks, such as authentication and identification attacks, availability attacks, confidentiality attacks, and data authenticity attacks. In this context, we provide a systematic review of emerging security and privacy vulnerabilities IoV and identify current challenges and the remaining open issues. Furthermore, we will discuss the future of this area. The paper contributes by presenting significant issues and solutions in a clear, smooth, comprehensive, and integrated way
Efficient Federated Intrusion Detection in 5G Ecosystem Using Optimized BERT-Based Model
No full textInternational audienceThe fifth-generation (5G) offers advanced services, supporting applications such as intelligent transportation, con-nected healthcare, and smart cities within the Internet of Things (IoT). However, these advancements introduce significant security challenges, with increasingly sophisticated cyber-attacks. This paper proposes a robust intrusion detection system (IDS) using federated learning and large language models (LLMs). The core of our IDS is based on BERT, a transformer model adapted to identify malicious network flows. We modified this transformer to optimize performance on edge devices with limited resources. Experiments were conducted in both centralized and federated learning contexts. In the centralized setup, the model achieved an inference accuracy of 97.79 %. In a federated learning context, the model was trained across multiple devices using both IID (Independent and Identically Distributed) and non-IID data, based on various scenarios, ensuring data privacy and compliance with regulations. We also leveraged linear quantization to com-press the model for deployment on edge devices. This reduction resulted in a slight decrease of 0.02 % in accuracy for a model size reduction of 28.74 %. The results underscore the viability of LLMs for deployment in IoT ecosystems, highlighting their ability to operate on devices with constrained computational and storage resources
Dynamic Coalition Formation among IoT Service Providers: A Systematic Exploration of IoT Dynamics Using an Agent-Based Model
No full textInternational audienceThis paper introduces an Agent-Based Model (ABM) designed to investigate the dynamics of the Internet of Things (IoT) ecosystem, focusing on dynamic coalition formation among IoT Service Providers (SPs). Drawing on insights from our previous research in 5G network modeling, the ABM captures intricate interactions among devices, Mobile Network Operators (MNOs), SPs, and customers, offering a comprehensive framework for analyzing the IoT ecosystem’s complexities. In particular, to address the emerging challenge of dynamic coalition formation among SPs, we propose a distributed Multi-Agent Dynamic Coalition Formation (MA-DCF) algorithm aimed at enhancing service provision and fostering collaboration. This algorithm optimizes SP coalitions, dynamically adjusting to changing demands over time. Through extensive experimentation, we evaluate the algorithm’s performance, demonstrating its superiority in terms of both payoff and stability compared to three classical coalition formation algorithms: static coalition, non-overlapping coalition, and random coalition. This study significantly contributes to a deeper understanding of the IoT ecosystem’s dynamics and highlights the potential benefits of dynamic coalition formation among SPs, providing valuable insights and opening future avenues for exploration
Segment Anything Model and Fully Convolutional Data Description for Plant Multi-Disease Detection on Field Images
No full textInternational audienceResearchers have designed various models trained on public or private datasets for plant disease detection to help farmers remedy crop yield losses on their farms due to plant diseases. Plantvillage is the most widely used plant disease dataset with laboratory images captured under controlled conditions with a single leaf on each image and a uniform background. Models trained on such datasets have extremely low classification accuracies when running on field images captured directly from plantations with various interwoven leaves, complex backgrounds, and different lighting conditions. In this study, we propose a model ensemble solution for the accurate identification and classification of plant diseases using field images. The model uses Segment Anything Model to efficiently circumscribe all identifiable objects in the image. Image Processing techniques are then used to isolate the identified objects from the original image. Background objects are separated from actual leaf objects using Fully Convolutional Data Description, which is an explainable deep one-class classification model for anomaly detection. Finally, the selected leaves are submitted to a Plantvillage-trained classification model for inference. Our model can detect diseases appearing on individual leaves of the same image and improves classification accuracy by more than 10% on public field plant disease datasets such as PlantDoc, thus providing a reliable solution for farmers and practitioners
Advancing UAV security with artificial intelligence: A comprehensive survey of techniques and future directions
No full textInternational audienceUnmanned Aerial Vehicles (UAVs) have become an integral part of modern smart cities and systems. However, the proliferation of UAVs has also brought a significant security concern. Therefore, UAVs security has attracted the interest of researchers in the field. To safeguard these leading devices, researchers and developers adopted artificial intelligence (AI) to improve the security system. Besides, emerging AI frameworks have entirely changed the landscape of UAVs quality of experience (QoE) by contributing to diverse issues. Meanwhile, owing to the novelty of the subject, AI-based frameworks for UAVs security were not carefully surveyed. The purpose of this study is to present a comprehensive survey of AI-based solutions for UAVs security provisioning through different techniques and algorithms. We assess and identify the recent AI solutions and the open research problems in the emerging field of UAV security. The study begins with an overview of UAV basics, their primary components, and the different communication links among networked UAVs. In addition, the paper explores UAVs security issues, threats, and attacks. This led to a discussion of proposed AI-based solutions for prevention and detection. Moreover, the paper categorizes AI-based solutions for UAV security and proposes a taxonomy to organize these approaches effectively. Finally, the review identifies open issues and future directions, offering valuable insights for researchers. The study methodology includes an extensive review of relevant publications in the past five years across UAVs, cyber-attacks, and AI. Therefore, this review presents an essential resource for understanding the current landscape of AI-based solutions for UAV security
An overview on the disassembly line balancing under uncertainty
No full textInternational audienceIn response to the growing emphasis on recycling and reusing end-of-life products, the effective management of disassembly processes under uncertainty has become a significant area of interest for researchers and practitioners alike. This paper offers an initial comprehensive review of the literature concerning the disassembly line balancing under uncertainty. The problem of stochastic line balancing is defined as the assignment of disassembly tasks to workstations in an uncertain environment while achieving specific objectives. Our review examines the diverse uncertainties discussed in the literature. We analyze the modeling techniques, approaches, and objectives associated with each type of uncertainty. Drawing from insights gained through the review of 54 publications, we subsequently engage in a discussion and outline potential directions for future research
Approximation of Superimposed Renewal Processes
No full textInternational audienceA Superimposed Renewal Process (SRP) consists in the observation of inter-arrival times formed by the superposition of multiple independent renewal processes. In imperfect maintenance analysis, a repairable series system is commonly modelled by an SRP when maintenance actions consist of renewing the failed components while leaving the other components unchanged. An SRP involves a large number of parameters and in practice, the identification of the renewed components is often unrecorded. Therefore, an SRP is often reduced to a simpler process, usually a Homogeneous Poisson Process. This paper presents two new approximations of an SRP, one based on virtual ages, the other based on copulas. Both approximation methods make it possible to model the dependency between two successive inter-arrival times and to obtain a precise assessment of the Remaining Useful Life. The performance of the approximation methods is discussed when the parameters of the models are assumed to be known
3D MRI Volume Segmentation Using 2D U-Net Models: A Focus on Deep Brain Structures of the Striatum for Parkinson's Disease early diagnosis
No full textInternational audienceParkinson's disease (PD) is a progressive neurodegenerative disorder characterized by motor symptoms, affecting millions of people worldwide. It results from the loss of dopamine in critical brain structures such as the substantia nigra and the striatum. Nowadays, several research studies highlight the importance of medical imaging techniques for early PD diagnosis. Among these techniques, Magnetic Resonance Imaging (MRI) holds a central position in neuroimaging, offering unparalleled insights into deep brain structures. Nevertheless, accurate segmentation of complex brain regions, such as the striatum, remain essential for precise diagnoses and decision-making. In this study, we present a novel method that utilizes 2D U-Net models to partition 3D MRI volumes and automatically segment the striatum, avoiding multi-class segmentation complexities. Our dataset comprised 97 MRI volumes with their respective segmentation. Each 3D MRI volume was converted into 3112D images slices of 128x128 dimensions. In our methodology, we applied two distinct data preprocessing techniques to generate optimal inputs for our U-N et model. We proposed both manual and automatic selection method to reduce the number of slices by filtering out homogeneous and redundant ones. Comprehensive evaluation metrics, including Dice Loss and Binary Cross Entropy, were employed to ensure rigorous assessment. Results demonstrated promising potential with filtered slices yielding better striatum segmentation and a training accuracy of 90.59%
Une métrique de maturité pour aligner les acteurs d'écosystèmes d'innovation en santé: Le cas des Concept Maturity Levels
No full textInternational audienceLes organisations du secteur de la santé font face à de nombreux défis cliniques, technologiques et réglementaires pour répondre aux besoins évolutifs des patients. Ces défis rendent la gestion des projets d’innovation complexe et difficile à appréhender pour une organisation seule, les poussant à adopter des stratégies d’écosystèmes. Pour fonctionner, ces écosystèmes d’innovation requièrent des efforts d’alignement entre des acteurs hétérogènes. Alors que la littérature sur l’alignement des acteurs s’est développée ces dernières années à l’échelle intra-organisationnelle, peu de recherches abordent la réalisation concrète de cet alignement à l’échelle inter-organisationnelle. Dans cet article, nous explorons le développement d’un tel alignement au sein d’écosystèmes d’innovation grâce aux métriques de maturité, telles que les Concept Maturity Levels. Notre analyse de plusieurs cas d’usage dans des écosystèmes d’innovation en santé en France révèle que les acteurs utilisent les Concept Maturity Levels pour cartographier, ordonner et normer leurs activités, compétences et ressources, permettant ainsi à leur alignement. Nous contribuons à la littérature des métriques de maturité en suggérant un usage des métriques de maturité plus extensif et exploratoire à l’échelle inter-organisationnelle, dû au nombre et à l’hétérogénéité d’acteurs impliqués. Nous contribuons également à la littérature sur les écosystèmes en suggérant des pratiques concrètes d’alignement des acteurs par l’usage de métriques de maturité. Enfin, nous présentons l’implication de ces résultats pour le secteur de la santé et les opportunités de recherches futures