Publications

2020

• Optimal probing sequences for polarization-multiplexed coherent phase OTDR
  
  • Dorize Christian
  • Awwad Elie
  • Guerrier Sterenn
  • Renaudier Jérémie
  , 2020, pp.T3.23. (10.1364/OFS.2020.T3.23)
  DOI : 10.1364/OFS.2020.T3.23
• Confidence-based Weighted Loss for Multi-label Classification with Missing Labels
  
  • Ibrahim Karim M
  • Epure Elena
  • Peeters Geoffroy
  • Richard Gael
  , 2020. The problem of multi-label classification with missing labels (MLML) is a common challenge that is prevalent in several domains, e.g. image annotation and auto-tagging. In multi-label classification, each instance may belong to multiple class labels simultaneously. Due to the nature of the dataset collection and labelling procedure , it is common to have incomplete annotations in the dataset, i.e. not all samples are labelled with all the corresponding labels. However, the incomplete data labelling hinders the training of classification models. MLML has received much attention from the research community. However, in cases where a pre-trained model is fine-tuned on an MLML dataset, there has been no straightforward approach to tackle the missing labels, specifically when there is no information about which are the missing ones. In this paper, we propose a weighted loss function to account for the confidence in each label/sample pair that can easily be incorporated to fine-tune a pre-trained model on an incomplete dataset. Our experiment results show that using the proposed loss function improves the performance of the model as the ratio of missing labels increases. (10.1145/3372278.3390728)
  DOI : 10.1145/3372278.3390728
• Towards Polarization-Insensitive Coherent Coded Phase OTDR
  
  • Guerrier Sterenn
  • Dorize Christian
  • Awwad Elie
  • Renaudier Jérémie
  , 2020, pp.T3.20. (10.1364/OFS.2020.T3.20)
  DOI : 10.1364/OFS.2020.T3.20
• VeReMi Extension: A Dataset for Comparable Evaluation of Misbehavior Detection in VANETs
  
  • Kamel Joseph
  • Wolf Michael
  • Heijden Rens Wouter van Der
  • Kaiser Arnaud
  • Urien Pascal
  • Kargl Frank
  , 2020. Cooperative Intelligent Transport Systems (C-ITS) is a new upcoming technology that aims at increasing road safety and reducing traffic accidents. C-ITS is based on peer-to-peer messages sent on the Vehicular Ad hoc NETwork (VANET). VANET messages are currently authenticated using digital keys from valid certificates. However, the authenticity of a message is not a guarantee of its correctness. Consequently, a misbehavior detection system is needed to ensure the correct use of the system by the certified vehicles. Although a large number of studies are aimed at solving this problem, the results of these studies are still difficult to compare, reproduce and validate. This is due to the lack of a common reference dataset. For this reason, the original VeReMi dataset was created. It is the first public misbehavior detection dataset allowing anyone to reproduce and compare different results. VeReMi is used in a number of studies and is currently the only dataset in its field. In this Paper, we extend the dataset by adding realistic a sensor error model, a new set of attacks and larger number of data points. Finally, we also provide benchmark detection metrics using a set of local detectors and a simple misbehavior detection mechanism. (10.1109/ICC40277.2020.9149132)
  DOI : 10.1109/ICC40277.2020.9149132
• Privacy-Preserving Data-Prefetching in VehicularNetworks via Reinforcement Learning
  
  • Berri Sara
  • Zhang Jun
  • Bensaou Brahim
  • Labiod Houda
  , 2020.
• Design Space Exploration with Deterministic Latency Guarantees for Crossbar MPSoC Architectures
  
  • Uscumlic Bogdan
  • Enrici Andrea
  • Pacalet Renaud
  • Gharbi Amna
  • Apvrille Ludovic
  • Natarianni Lionel
  • Roullet Laurent
  , 2020. MPSoC and NoC systems are often used in complex telecommunication systems, which in the 5G era need to enable telecommunication services with unprecedented latency characteristics. Indeed, new services emerge, needing deterministic latency guarantees with virtually no system jitter, during the lifetime of the established telecommunication service. In this work, for the first time, we propose an optimal solution for a design space exploration (DSE) optimization problem, that performs all the traditional DSE tasks, but with end-to-end deterministic latency guarantees. We focus on MPSoC or NoC architectures with crossbars, although this work can be easily extended to more complex architectures. More precisely, our contributions in this work are the following: 1) we propose a novel method for deterministic scheduling in MPSoC and NoC architectures with a crossbar; 2) we propose an optimal solution in the form of an integer linear program (ILP) for DSE problem with end-to-end deterministic latency guarantees; 3) we identify the trade-off between the latency due to the use of crossbar time slots and the application execution time at different processing elements. The numerical results suggest that the proposed deterministic scheduling method can efficiently use all 100% of the crossbar capacity, depending on available application load and system parameters.
• An Intelligent Mechanism for Sybil Attacks Detection in VANETs
  
  • Quevedo Carlos
  • Quevedo Ana
  • Campos Gustavo
  • Gomes Rafael
  • Celestino Joaquim
  • Serhrouchni Ahmed
  , 2020, pp.1-6. Vehicular Ad Hoc Networks (VANETs) have a strategic goal to achieve service delivery in roads and smart cities, considering the integration and communication between vehicles, sensors and fixed road-side components (routers, gateways and services). VANETs have singular characteristics such as fast mobile nodes, self-organization, distributed network and frequently changing topology. Despite the recent evolution of VANETs, security, data integrity and users privacy information are major concerns, since attacks prevention is still open issue. One of the most dangerous attacks in VANETs is the Sybil, which forges false identities in the network to disrupt compromise the communication between the network nodes. Sybil attacks affect the service delivery related to road safety, traffic congestion, multimedia entertainment and others. Thus, VANETs claim for security mechanism to prevent Sybil attacks. Within this context, this paper proposes a mechanism, called SyDVELM, to detect Sybil attacks in VANETs based on artificial intelligence techniques. The SyDVELM mechanism uses Extreme Learning Machine (ELM) with occasional features of vehicular nodes, minimizing the identification time, maximizing the detection accuracy and improving the scalability. The results suggest that the suitability of SyDVELM mechanism to mitigate Sybil attacks and to maintain the service delivery in VANETs. (10.1109/ICC40277.2020.9149371)
  DOI : 10.1109/ICC40277.2020.9149371
• Physics and applications of quantum dot lasers for silicon photonics
  
  • Grillot Frédéric
  • Norman Justin
  • Duan Jianan
  • Zhang Zeyu
  • Dong Bozhang
  • Huang Heming
  • Chow Weng
  • Bowers John
  Nanophotonics, De Gruyter, 2020, 9 (6), pp.1271-1286. (10.1515/nanoph-2019-0570)
  DOI : 10.1515/nanoph-2019-0570
• Improving IoT data stream analytics using summarization techniques
  
  • Bahri Maroua
  , 2020. With the evolution of technology, the use of smart Internet-of-Things (IoT) devices, sensors, and social networks result in an overwhelming volume of IoT data streams, generated daily from several applications, that can be transformed into valuable information through machine learning tasks. In practice, multiple critical issues arise in order to extract useful knowledge from these evolving data streams, mainly that the stream needs to be efficiently handled and processed. In this context, this thesis aims to improve the performance (in terms of memory and time) of existing data mining algorithms on streams. We focus on the classification task in the streaming framework. The task is challenging on streams, principally due to the high -- and increasing -- data dimensionality, in addition to the potentially infinite amount of data. The two aspects make the classification task harder.The first part of the thesis surveys the current state-of-the-art of the classification and dimensionality reduction techniques as applied to the stream setting, by providing an updated view of the most recent works in this vibrant area.In the second part, we detail our contributions to the field of classification in streams, by developing novel approaches based on summarization techniques aiming to reduce the computational resource of existing classifiers with no -- or minor -- loss of classification accuracy. To address high-dimensional data streams and make classifiers efficient, we incorporate an internal preprocessing step that consists in reducing the dimensionality of input data incrementally before feeding them to the learning stage. We present several approaches applied to several classifications tasks: Naive Bayes which is enhanced with sketches and hashing trick, k-NN by using compressed sensing and UMAP, and also integrate them in ensemble methods.
• Surrogate modeling of stochastic simulators
  
  • Azzi Soumaya
  , 2020. This thesis is a contribution to the surrogate modeling and the sensitivity analysis on stochastic simulators. Stochastic simulators are a particular type of computational models, they inherently contain some sources of randomness and are generally computationally prohibitive. To overcome this limitation, this manuscript proposes a method to build a surrogate model for stochastic simulators based on Karhunen-Loève expansion. This thesis also aims to perform sensitivity analysis on such computational models. This analysis consists on quantifying the influence of the input variables onto the output of the model. In this thesis, the stochastic simulator is represented by a stochastic process, and the sensitivity analysis is then performed on the differential entropy of this process.The proposed methods are applied to a stochastic simulator assessing the population’s exposure to radio frequency waves in a city. Randomness is an intrinsic characteristic of the stochastic city generator. Meaning that, for a set of city parameters (e.g. street width, building height and anisotropy) does not define a unique city. The context of the electromagnetic dosimetry case study is presented, and a surrogate model is built. The sensitivity analysis is then performed using the proposed method.
• Improved Optimistic Algorithms for Logistic Bandits
  
  • Faury Louis
  • Abeille Marc
  • Calauzènes Clément
  • Fercoq Olivier
  , 2020. The generalized linear bandit framework has attracted a lot of attention in recent years by extending the well-understood linear setting and allowing to model richer reward structures. It notably covers the logistic model, widely used when rewards are binary. For logistic bandits, the frequentist regret guarantees of existing algorithms areÕ(κ √ T), where κ is a problem-dependent constant. Unfortunately, κ can be arbitrarily large as it scales exponentially with the size of the decision set. This may lead to significantly loose regret bounds and poor empirical performance. In this work, we study the logistic bandit with a focus on the prohibitive dependencies introduced by κ. We propose a new optimistic algorithm based on a finer examination of the non-linearities of the reward function. We show that it enjoys aÕ(√ T) regret with no dependency in κ, but for a second order term. Our analysis is based on a new tail-inequality for self-normalized martingales, of independent interest.
• Resource allocation for latency sensitive wireless systems
  
  • Avranas Apostolos
  , 2020. The new generation of wireless systems 5G aims not only to convincingly exceed its predecessor (LTE) data rate but to work with more dimensions. For instance, more user classes were introduced associated with different available operating points on the trade-off of data rate, latency, reliability. New applications, including augmented reality, autonomous driving, industry automation and tele-surgery, push the need for reliable communications to be carried out under extremely stringent latency constraints. How to manage the physical level in order to successfully meet those service guarantees without wasting valuable and expensive resources is a hard question. Moreover, as the permissible communication latencies shrink, allowing retransmission protocol within this limited time interval is questionable. In this thesis, we first pursue to answer those two questions. Concentrating on the physical layer and specifically on a point to point communication system, we aim to answer if there is any resource allocation of power and blocklength that will render an Hybrid Automatic ReQuest (HARQ) protocol with any number of retransmissions beneficial. Unfortunately, the short latency requirements force only a limited number of symbols to possibly be transmitted which in its turn yields the use of the traditional Shannon theory inaccurate. Hence, the more involved expression using finite blocklength theory must be employed rendering the problem substantially more complicate. We manage to solve the problem firstly for the additive white gaussian noise (AWGN) case after appropriate mathematical manipulations and the introduction of an algorithm based on dynamic programming. Later we move on the more general case where the signal is distorted by a Ricean channel fading. We investigate how the scheduling decisions are affected given the two opposite cases of Channel State Information (CSI), one where only the statistical properties of the channel is known, i.e. statistical CSI, and one where the exact value of the channel is provided to the transmitter, i.e., full CSI.Finally we ask the same question one layer above, i.e. the Medium Access Contron (MAC). The resource allocation must be performed now accross multiple users. The setup for each user remains the same, meaning that a specific amount of information must be delivered successfully under strict latency constraints within which retransmissions are allowed. As 5G categorize users to different classes users according to their needs, we model the traffic under the same concept so each user belongs to a different class defining its latency and data needs. We develop a deep reinforcement learning algorithm that manages to train a neural network model that competes conventional approaches using optimization or combinatorial algorithms. In our simulations, the neural network model actually manages to outperform them in both statistical and full CSI case.
• Improved Virtual Anchor Selection for AR-assisted Sensor Positioning in Harsh Indoor Conditions
  
  • Choi Hong-Beom
  • Lim Keun-Woo
  • Ko Young-Bae
  , 2020, pp.1-6. In this work, we propose a sensor localization system assisted by wireless communication and augmented reality (AR) suitable for harsh indoor environments. Future handheld and unattended devices will be equipped with technologies, which have the means to enable localization of various phenomenon in indoor environments. These include visual odometry based on cameras and augmented reality, and communication hardware such as UWB. Integration of such technologies allows us to compensate for each other’s errors in measurements. However, existing work cannot fully exploit these technologies to high extent, often inducing more errors or wasted resources. Furthermore, many other existing indoor localization methods require a pre-defined architecture or building information, which are impossible to acquire in harsh indoor conditions. In our proposed system, we specifically propose an improved method of utilizing virtual anchors for localization of sensor tags in harsh indoor environments, based on a cluster-based selection algorithm that allows the system to improve the positioning accuracy through angular diversity while also improving resource utilization without existing knowledge of the location. Our work has been fully implemented and tested on several indoor environments. (10.1109/GIOTS49054.2020.9119679)
  DOI : 10.1109/GIOTS49054.2020.9119679
• Sequence-to-Sequence Predictive models: from Prosody to Communicative Gestures
  
  • Yunus Fajrian
  • Clavel Chloé
  • Pelachaud Catherine I
  , 2020. Communicative gestures and speech prosody are tightly linked. Our aim is to predict when gestures are performed based on prosody. We develop a model based on a seq2seq recurrent neural network with attention mechanism. The model is trained on a corpus of natural dyadic interaction where the speech prosody and the gestures have been annotated. Because the output of the model is a sequence, we use a sequence comparison technique to evaluate the model performance. We find that the model can predict certain gesture classes. In our experiment, we also replace some input features with random values to find which prosody features are pertinent. We find that the F0 is pertinent. Lastly, we also train the model on one speaker and test it with the other speaker to find whether the model is generalisable. We find that the models which we train on one speaker also works for another speaker of the same conversation.
• End-to-end automated cache-timing attack driven by machine learning
  
  • Perianin Thomas
  • Carré Sebastien
  • Dyseryn Victor
  • Facon Adrien
  • Guilley Sylvain
  Journal of Cryptographic Engineering, Springer, 2020, 11 (2), pp.135-146. Cache-timing attacks are serious security threats that exploit cache memories to steal secret information. We believe that the identification of a sequence of function calls from cache-timing data measurements is not a trivial step when building an attack. We present a recurrent neural network model able to automatically retrieve a sequence of operations from cache-timings. Inspired from natural language processing, our model is able to learn on partially labelled data. We use the model to unfold an end-to-end automated attack on OpenSSL ECDSA on the secp256k1 curve. Our attack is able to extract the 256 bits of the secret key by auto- matic analysis of about 2400 traces without any human processing. (10.1007/s13389-020-00228-5)
  DOI : 10.1007/s13389-020-00228-5
• Multi-hop Data Fragmentation in Energy Harvesting Wireless Sensor Networks
  
  • Lim Keun-Woo
  • Kapusta Katarzyna
  • Memmi Gerard
  • Jung Woo-Sung
  , 2020, pp.1-6. In this work, we utilize multi-hop data fragmentation in energy harvesting wireless sensor networks (EHWSN) for improved security and routing. 1 Data fragmentation is a secure data management and data provisioning technique used in unattended sensor networks. However, it compromises energy efficiency for better protection, possibly decreasing the lifetime of a traditional sensor network. Instead, it has potentials to be used in EHWSN, which allows better energy utilization for each sensor node while security of sensed data remains an important factor. In our work, we consider how data fragmentation can be applied in EHWSN, and propose a multi-hop data fragmentation method suitable for EHWSN (10.1109/GIOTS49054.2020.9119569)
  DOI : 10.1109/GIOTS49054.2020.9119569
• Analyse multimodale de la cohésion de groupe
  
  • B Kantharaju Reshmashree
  • Langlet Caroline
  • Barange Mukesh
  • Clavel Chloé
  • Pelachaud Catherine I
  , 2020.
• Enacting Critical Performativity: Normative And Empirical Foundations For A Critical Engagement
  
  • Ouahab Alban
  • Acquier Aurélien
  , 2020.
• Guider l'attention dans les modèles de séquence à séquence pour la prédiction des actes de dialogue
  
  • Chapuis Emile
  • Colombo Pierre
  • Manica Matteo
  • Varni Giovanna
  • Vignon Emmanuel
  • Clavel Chloé
  , 2020. La prédiction d’actes de dialogue (AD) basés sur le dialogue conversationnel est un élément clé dans le développement des agents conversationnels. La prédiction précise des AD nécessite une modélisation précise à la fois de la conversation et des dépendances globales des AD. Nous utilisons les approches de séquence à séquence (seq2seq) largement adoptées dans la traduction automatique neurale (NMT) pour améliorer la modélisation de la séquentialité des AD. Les modèles seq2seq sont connus pour apprendre les dépendances globales complexes alors que les approches actuellement proposées utilisant des champs aléatoires conditionnels linéaires (CRF) ne modélisent que les dépendances locales des AD. Dans ce travail, nous introduisons un modèle seq2seq adapté à la classification AD en utilisant : un codeur hiérarchique, un nouveau mécanisme attention guidée et la recherche de faisceau appliquée à la fois à l’apprentissage et à l’inférence. Par rapport à l’état de l’art, notre modèle ne nécessite pas de caractéristiques artisanales et est formé de bout en bout. En outre, l’approche proposée obtient un score de précision inégalé de 85% pour la SwDA et un score de précision de pointe de 91,6% pour la MRDA.
• Tu sais qui sait quoi ? Suggestions pour l'étude du système de mémoire transactive dans un groupe à partir des patterns comportementaux et conversationnels.
  
  • Biancardi Beatrice
  • Maisonnave-Couterou Lou
  • Mancini Maurizio
  • Varni Giovanna
  , 2020. Le Système de Mémoire Transactive (TMS) représente la division coopérative du travail permettant au groupe d’apprendre. Les processus psychologiques sous-tendant le TMS ont fait l’objet de nombreuses études, pourtant peu de chercheurs se sont interrogés sur la pertinence des indices comportementaux dans la prédiction du niveau de TMS au sein d’un groupe. Cet article suit alors deux objectifs : la présentation d’une méthodologie permettant la reproduction des différentes phases du TMS (i.e. encodage, stockage et récupération) et le recueil des données concernant les interactions d’un groupe pendant une tâche commune. Par ailleurs, nous relevons ces données à partir de mesures objectives (i.e. F-formation, distance interpersonnelle, tour de parole) et subjectives (i.e. auto-évaluation). Ainsi, par l’intermédiaire de notre étude, nous tentons de relever des patterns spatiaux et conversationnels pertinents dans la prédiction du niveau de TMS d’un groupe. Nous cherchons notamment à comprendre comment ces différents patterns pourraient être liés aux trois dimensions majeures du TMS (i.e. spécialisation des connaissances, confiance et coordination). Par la suite, nos résultats devraient pouvoir contribuer aux développements de modèles computationnels, mais aussi d’applications fonctionnant en temps réel et pouvant être exploitées dans plusieurs contextes (e.g. équipes chirurgicales, team design, cours en ligne).
• Temperature dependent linewidth rebroadening in quantum dot semiconductor lasers
  
  • Köster Felix
  • Duan Jianan
  • Dong Bozhang
  • Grillot Frédéric
  • Lüdge Kathy
  Journal of Physics D: Applied Physics, IOP Publishing, 2020, 53 (23), pp.235106. (10.1088/1361-6463/ab7ca5)
  DOI : 10.1088/1361-6463/ab7ca5
• GRACE : Un projet portant sur l'étude automatique de la cohésion dans les petits groupes d'humains
  
  • Maman Lucien
  • Varni Giovanna
  , 2020. Cet article présente le projet GRACE (GRoups’ analysis for Automated Cohesion Estimation), un projet de recherche fondamentale JCJC financé par l’Agence Nationale de la Recherche française qui vise à développer un modèle informatique de la cohésion dans les interactions humain-humain. Premièrement, les objectifs du projet sont décrits. Ensuite, une brève revue de l’état de l’art sur la cohésion est détaillée. Enfin, la méthodologie adoptée est présentée.
• How Confident are You? Exploring the Role of Fillers in the Automatic Prediction of a Speaker's Confidence
  
  • Dinkar Tanvi
  • Vasilescu Ioana
  • Pelachaud Catherine
  • Clavel Chloé
  , 2020. “Fillers", example “um" in English, have been linked to the “Feeling of Another’s Knowing (FOAK)" or the listener’s perception of a speaker’s expressed confidence. Yet, in Spoken Language Processing (SLP) they remain unexplored, or overlooked as noise. We introduce a new challenging task for educational applications, that is the prediction of FOAK. We design a set of filler features based on linguistic literature, and investigate their potential in FOAK prediction. We show that the integration of information related to implicature meanings allows an improvement in the FOAK model and that the different functions of fillers are differently correlated with confidence.
• Attention Slices dans les Entretiens d'Embauche Vidéo Différés
  
  • Hemamou Léo
  • Guillon Arthur
  • Martin Jean-Claude
  • Clavel Chloé
  , 2020. Dans cet article, nous nous intéressons à l’étude de signaux influents dans les entretiens vidéo d’embauche asynchrones découverts par des méthodes d’apprentissage profond. Le système que nous étudions emploie des mécanismes d’attention, qui permettent d’extraire d’un entretien les informations et les instants décisifs (qui ont influencé la décision du système au niveau de l’entretien), sans requérir d’annotation locale. Alors que la majorité des approches similaires évaluent les mécanismes d’attention en se contentant de visualiser les moments d’attention maximale, nous proposons ici une méthodologie permettant d’automatiser l’analyse du contenu de ces attention slices afin de fournir des éléments d’interprétation des prédictions du système.
• Proving IoT Devices Firmware Integrity With Bijective MAC Time Stamped
  
  • Urien Pascal
  , 2020, pp.1-2. (10.1109/WF-IoT48130.2020.9221395)
  DOI : 10.1109/WF-IoT48130.2020.9221395