Research

Habilitation defences

«Traitement statistique pour les communications numériques et les images»

Who ?

The habilitation will be defended by Frédéric Lehmann, Assistant Professor in CITI Department

When ?

Friday 10th of February, 2012 at 2.00 PM

Where ?

At Telecom SudParis, 9 Rue Charles Fourier, 91000 Evry - Amphi 10

Board

• M. Jean-Yves TOURNERET, Professor, INP - ENSEEIHT, Toulouse - reporter
• M. Ramesh PYNDIAH, Professor, Telecom Bretagne, Brest - reporter
• M. Benoît GELLER, Professor, ENSTA, Paris - reporter
• M. Jean-Luc ZARADER, Professor, ISIR, UPMC, Paris - examiner
• M. Jean-Pierre CANCES, Professor, ENSIL, Limoges - examiner
• M. Antoine BERTHET, Professor, SUPELEC, Gif-sur-Yvette - examiner

---

«Sécurité, vie privée et confiance dans les environnements pervasifs»

Who ?

The habilitation was defended by Abdallah Mhamed, Assistant Professor in RST Department

When ?

Thursday 30th of June, 2011 at 1.30 PM

Where ?

At Telecom SudParis, 9 Rue Charles Fourier, 91000 Evry - room A003

Board

• M. Gildas Avoine, Professor, Information Security Group, Louvain la Neuve, Belgique - reporter
• M. Lionel Brunie, Professor, LIRIS, INSA, Villeurbanne, France - reporter
• M. Sébastien Canard, Professor, France Telecom R&D, Caen , France - reporter
• Miss Amal El Fellah Seghrouchni, Professor, LIP6 , UPMC, Paris , France - examiner
• M. Jacques Demongeot, Professor, TIMC-IMAG, La Tronche, France - examiner

Abstract

"Ubiquitous computing and pervasive systems allows the development of devices and services improving the quality of life of dependant users. Nevertheless, due to the heterogeneity and the complexity of deployed equipments/services, it is important to consider the safety and security issues for dependant users living in these environments. To satisfy the specific needs of dependant users, it is necessary to consider the dynamic aspect of security in these intelligent environments by developing new security models which are able to adapt to different user profiles, user contexts and user environments. Mobile Ad-hoc networks (MANET) contribute significantly to the deployment of services in pervasive environments. Despite the huge research work devoted to some security services for MANETs , their consequences in terms of user privacy and trust are still not enough taken into consideration.

The aim of my research work is to cover the tightly related issues of security, privacy and trust in order to be integrated in a common architecture adapted to intelligent environments and dedicated for dependant users. To take into account the variability and the randomness of these environments, we should introduce the concept of « context awareness » which links the user profile (abilities, preferences), the time and the location (living space, service platform, devices) to provide personalized services.

Owing to various contributions on cryptographic protocols, have reached the design and the implementation of an authentication framework integrating security, privacy and trust in context aware environments."

---

«Model-based Conformance and Interoperability Testing»

Who ?

The habilitation was defended by Stéphane Maag, Assistant Professor in LOR Department

When ?

Wednesday 18th of May, 2011 at 5.00 PM

Where ?

At Paris Sud XI University - Bat. 490, room 79, Rue du Doyen Joseph Pérés, 91405 Orsay.

Board

• M. Richard Castanet, Professor, LaBRI, Bordeaux I University, reporter
• M. Teruo Higashino, Professor, Osaka University, reporter
• M. Patrick Senac, Professor, ISAE, CNRS LAAS, reporter
• M. Khaldoun Alagha, Professor, LRI, Paris Sud XI University, examiner
• Miss Ana Cavalli, Professor, Telecom SudParis, examiner
• M. Philippe Dague, Professor, LRI, Paris Sud XI University, examiner
• M. Manuel Nunez, Professor, Complutense University, Madrid, examiner

Abstract

"Testing is mainly the process of checking that a system possesses a set of desired properties and behavior. It has become an integral part of innovation, production and operation of systems in order to reduce the risk of failures and to guarantee the quality and reliability of the software or protocol used. Through the efforts provided by the standardization institutes to normalize the testing activities, there is a high level of consciousness of its importance and impact for the future deployment and use of software and systems.

Today, there are different testing activities (such as unit testing, regression testing, etc.) for different kind of components or systems dedicated to the assessment of, for instance, their reliability, performance, robustness through the validation or verification processes.

Besides, there is an awareness of (i) an increasing need to automate the testing activity in all application areas, (ii) that this activity should be taken into consideration in all phases of the system life cycle, and (iii) that interoperability and standardization rely on it.

Different testing techniques may be performed. There are commonly two main ways to address testing activities: white box and black box testing techniques. These techniques depend on the level of accessibility we may have on the implementations. While we suppose that the source code and details of the system are available when we apply white box testing techniques, it is unavailable with black box ones. When testing a system in a black box context, a formal or even semi-formal model of that system is convenient especially to derive test suites checking one specific property of the system. The automatic and semi-automatic test suites derivation mechanisms are therefore detailed. We provide interesting assessments of the empirical existing techniques as well as application of more recent ones. Their advantages and drawbacks are studied. In my work, I am interested in model-based testing techniques and their studies and applications to the conformance and interoperability of communicating systems.

The different phases and issues of protocol and application modeling are discussed, and the test suites derivation and their application to provide usable verdicts finding out errors are analyzed."

---

«Aggregation of traffic and Markov chain models for QoS of complex networks»

Who ?

The habilitation was defended by Hind Castel, Assistant Professor in RST Department

When ?

Thursday 12th of May, 2011 at 2.00 PM

Where ?

At Paris VI University - Pierre et Marie Curie, Campus Jussieu, 4 place Jussieu Paris 5ème, Maison de la Pédagogie, first floor, room 25-26 101

Board

• M. Serge Haddad, Professor, LSV - ENS Cachan, reporter
• M. Tadeusz Czachorski, Research Dean, Polish Academy of Sciences, reporter
• M. Gérardo Rubino, Research Dean, INRIA/IRISA Rennes, reporter
• M. Serge Fdida, Professor, LIP6 - Paris 6, examiner
• M. Jean-Michel Fournau, Professor, PRiSM - Versailles Saint Quentin University, examiner
• M. Dominique Chiaroni, Research Engineer, Alcatel Lucent Labs, examiner
• Miss Nihal Pekergin, Professor, Paris 12, examiner

Abstract

"Today, networks are increasingly complex in terms of numbers of nodes, diversity of flow (flow, packet sizes, variability of traffic) or kinds of protocols. The guarantee of QoS (Quality of Service) remains an important and difficult problem to solve.

Two important aspects affecting the QoS have been studied: traffic management, and methods of performance evaluation based on stochastic bounds. Aggregation is a very intuitive mechanism and very used in communication networks. It has been studied in this work in quite different areas: traffic, and Markov chains models. In both cases the idea is similar: we aggregate entities (packets or states) in order to generate a simplified system.

Different algorithms for traffic aggregation taking into account QoS of IP flows are proposed to define optical packets. The idea is to improve the use of the bandwidth of the optical fiber, and thus the network performance.

In performance analysis methods, aggregation algorithms of multidimensional Markov chains are proposed. The relevance is to build a Markov chain with a lower size from which we can calculate performance measures bounds. The aggregation algorithm  can offer a trade off between quality and complexity of the bound computations. Moreover, both stationary and transient measures can be derived.

It is clear that the aggregation of traffic and the aggregation of Markov systems are different techniques. However, they both contribute efficiently to QoS in communication networks. "

---

«Intergiciels pour la sensibilité au contexte en environnement ubiquitaire»

Who ?

The habilitation was defended by Chantal Taconet, Assistant Professor in INF Department

When ?

Thursday 10th of February, 2011 at 01.30 PM

Where ?

At Telecom SudParis, 9 Rue Charles Fourier, 91000 Evry - room C06

Board

• Yolande Berbers, Professor, Université catholique de Louvain, reporter
• Michel Riveill, Professor, École Polytechnique de l'Université de Nice, reporter
• Lionel Seinturier, Professor, Université de Lille, reporter
• Philippe Lalanda, Professor, Université de Grenoble, examiner
• Isabelle Demeure, Professor, Telecom ParisTech, examiner
• Guy Bernard, Professor, Telecom SudParis, examiner

Abstract

"En informatique ubiquitaire, les interactions des applications avec le monde extérieur changent de nature : aux interactions classiques (principalement des interactions avec les utilisateurs) s'ajoutent des interactions avec l'environnement extérieur. L'application doit alors se doter d'une capacité d'observation de son environnement. Le spectre des éléments potentiellement observables est large : de l'application elle-même, au dispositif mobile d'exécution, aux capteurs ambiants, aux objets communicants accessibles à grande échelle avec l'Internet des objets. L'application doit également avoir la capacité de réagir et de se reconfigurer en fonction des situations. Pour généraliser et faciliter la création de systèmes ubiquitaires en environnement réparti, il est nécessaire de fournir de nouveaux outils de conception ainsi que de nouveaux intergiciels. Ceci est d'autant plus important que la palette des applications ambiantes et ubiquitaires croît tant dans les domaines de la santé, de la domotique, du transport, que des réseaux sociaux.

Ce manuscrit présente les travaux que j'ai réalisés au cours des dix dernières années dans le domaine de l'intergiciel pour les applications réparties en environnement ubiquitaire. J'ai eu pour objectif de travailler sur la préoccupation de sensibilité au contexte et sa prise en charge par les intergiciels. La ligne conductrice de ces travaux a été de réunir les travaux IDM (ingénierie dirigée par les modèles) et intergiciels pour faciliter la prise en compte de la sensibilité au contexte des applications depuis leur conception jusqu'à leur exécution. Mes contributions portent sur le processus de conception d'applications ubiquitaires, sur les méta-modèles spécifiques pour la conception de modèles d'applications ubiquitaires et sur les intergiciels pour l'ubiquitaire. Les intergiciels s'appuient sur des modèles de sensibilité au contexte tant pendant la phase de déploiement que pendant la phase d'exécution de l'application. Les modèles de conception sont utilisés non seulement pour la production de code mais aussi à l'exécution. Ainsi, l'intergiciel peut par réflexivité connaître et gérer l'influence de l'environnement sur l'application. La présence des modèles à l'exécution permet de plus la reconfiguration de la sensibilité au contexte des applications pendant leur exécution.

Ces diverses contributions ont montré l'apport de l'IDM pour la conception et l'exécution d'applications ubiquitaires. Pour la suite, il me semble important d'étendre ces travaux sur un environnement grande échelle notamment pour définir une sensibilité au contexte prenant en compte des observations en provenance de l'Internet des objets, pour réaliser des adaptations réparties d'applications réparties et pour assurer le passage à l'échelle des mécanismes pour un nombre important d'utilisateurs. L'introduction de mécanismes de vérification des délais d'identification de situations et de mise en œuvre des adaptations me semble également tout à fait incontournable pour fournir des intergiciels pour l'ubiquitaire qui assurent des délais d'adaptation acceptables."

---

«On the IP model defying wireless, mobility, heterogeneity and ubiquity»

Who ?

The habilitation was defended by Hakima Chaouchi, Assistant Professor in LOR Department

When ?

Monday 13rd of December, 2010 at 11.00 AM

Where ?

At Faculté des Sciences d'Orsay, Université Paris Sud XI - 490 Rue du Doyen Joseph Pérés, 91405 Orsay - room 101

Board

• Georg Carle, Professor, Université Munich TUM, Allemagne - Reporter
• Christian Bonnet, Professor, Eurecom Sophia Antipolis, France - Reporter
• Jozef Wozniak, Université de Gdansk, Pologne - Reporter
• Vania Conan, Research Director, Thales Group, France - Examineur
• Hikmet Sari, Professor SUPELEC Gif-sur-Yvette, France - Examineur
• Khaldoun Al Agha, Professor, Université Paris XI, France - Examineur

---

«Mesures de qualité et fusion d'informations pour la vérification biométrique d'identité»

Who ?

The habilitation was defended by Sonia GARCIA, Assistant Professor in EPH Department

When ?

Fryday 3rd of December, 2010 at 2.00 PM

Where ?

At University Paris VI - Pierre et Marie Curie, Campus Jussieu, 4 place Jussieu Paris 5ème, Tower 55-65, 2nd floor, room 211

Board

• Mr Kevin BOWYER (University of Notre-Dame, USA), Reporter
• Mr Robert SABOURIN (Ecole de Technologie Supérieure de Montréal, Canada), Reporter
• Miss Nicole VINCENT (Université Paris V, France), Reporter
• Miss Bernadette DORIZZI (Telecom SudParis, France), examineur
• Mr Maurice MILGRAM (Université Paris VI, France), examineur
• Mr Javier ORTEGA-GARCIA (Universidad Autonoma de Madrid, Espagne), examineur

---

«Méthodes et systèmes de représentation visuelle»

Who ?

The habilitation was defended by Titus Zaharia, Assistant Professor in ARTEMIS Department

When ?

Monday 22nd of November, 2010 at 9.30 AM

Where ?

At University Paris VI - Pierre et Marie Curie, Campus Jussieu, 4 place Jussieu Paris 5ème, Building 41, 2nd floor, room 203-205

Board

• Mr Patrick GALLINARI Professor, UPMC, President of Board 
• Mr Claude LABIT, Research Director, INRIA, Reporter 
• Mr Rémy PROST, Professor, INSA Lyon, Reporter
• Mr Jean-Marie BECKER, Professor, CPE Lyon, Reporter
• Miss Françoise PRETEUX, Professor, Mines Paristech, Examineur
• Mr François GASPARD, Doctor, CEA-LIST, Examineur
• Mr Gérard MOZELLE, Doctor, Thomson Broadcast, Examineur

---

« Vers des systèmes d'informations adaptatifs, sémantiques et pervasifs : application aux EIAH »

Who ?

The habilitation was defended by Amel Bouzeghoub, Associate Professor in INF Department

When ?

Monday 7th of June, 2010 at 2.00 PM

Where ?

Université Paris 7 Diderot, Campus Jussieu, 2 Place Jussieu, 75005 Paris, Maison de la Pédagogie, Salle 202, Couloir B - 2ème étage.

Board

• Pr. Florence Sèdes, Université Paul Sabatier, Toulouse - Reporter
• Pr. Agathe Merceron, Beuth University of Applied Sciences, Berlin - Reporter
• Pr. Jérôme Gensel, Université Pierre Mendès France, Grenoble - Reporter
• Pr. Anne Doucet, Université Pierre & Marie Curie - Examineur
• Pr. Serge Garlatti, Télécom Bretagne - Examineur
• Pr. Bruno Defude Professeur, Télécom SudParis - Examineur

Abstract

"Les systèmes d'information adaptatifs, sémantiques et pervasifs sont conçus pour être accessibles sur Internet, possèdent une architecture distribuée, accèdent à une masse de ressources hétérogènes qu'il s'agit de réutiliser, de partager, assurent une adaptation/personnalisation en fonction des besoins des utilisateurs et de leur contexte et permettent la coopération homme/machine et/ou entre individus appartenant à une même communauté de pratique à travers le web.

Ces systèmes sont basés sur une boucle perception de l'environnement, découverte/sélection, décision, adaptation, déploiement. Cette problématique recouvre un large spectre d'applications. Les EIAH constituent un domaine d'application privilégié qui illustre une grande partie de cette problématique. En effet, les EIAH sont des environnements informatique conçus pour favoriser l'apprentissage humain, c'est-à-dire la construction ou le développement de connaissances chez un apprenant.

Nous présentons ici nos principaux résultats qui s'articulent autour de trois axes :

1. La composition adaptative de ressources;
2. L'adaptation au contexte et son application à l'apprentissage pervasif;
3. L'intégration sémantique de ressources hétérogènes.

---

«Natural information sources: from mathematical modelling to multimedia applications»

Who ?

The habilitation was defended by Mihai Petru Mitrea, Associate Professor in ARTEMIS Department

When ?

Wednesday 2nd of June, 2010 at 2.00 PM

Where ?

At "Université Paris VI - Pierre et Marie Curie, Campus Jussieu", room 211 (55-65)

Board

• Pr. Fréderic TRUCHETET, Université de Bourgogne, Reporter
• Pr. Touradj EBRAHIMI, Ecole Polytechnique Fédérale de Lausanne, Reporter
• Pr. Michel JOURLIN, Université de Saint Etienne, Reporter
• Pr. Patrick GALLINARI, Université Paris VI, board member
• Pr. Françoise PRETEUX, MINES ParisTech, board member
• M. Philippe MELLE, SFR, board member

Abstract

"The overall research, contractual, standardisation and educational activities presented in this HDR can be structured along three main axes: (1) Mathematical analysis of natural information sources, (2) Content enrichment and (3) Exchange and interaction.

Mathematical analysis of the natural information sources

The main direction of this study is to obtain accurate statistical models (i.e. probability estimation with error control and the underlying stationarity/ergodicity investigation) for several natural information source intensively considered in the nowadays applications: written language, video, and the multimedia content noise.
Note that regardless its final application (compression, indexing, watermarking, content enrichment, to mention but a few), each and every time a new multimedia processing method is developed, the stationarity is always assumed. Unfortunately, because of the very mathematical meaning of this concept, the literature is scarce concerning the viability of this hypothesis. This gap is bridged in the present HDR by advancing novel statistical approaches granting stationaritry/ergodicity support in the estimation procedure; in this respect, both traditional and new developed statistical tools are combined. The final results consist in reference models for the three above mentioned information sources:

• Natural language: The models consist in 95% confidence intervals for letter/digram/trigram/tetragram/word probabilities. The accuracy of the model is expressed by the related relative errors in probability estimation. The stationarity/ergodicity proof results from the developed investigation procedure and is expressed by the significance levels in statistical tests (5% for both test on the equality between two probabilities and test on the hypothesis that probability belongs to an interval) and in the size of the type two statistical error in the test on the hypothesis that probability belongs to an interval (lower than 15%). Note that the test on the hypothesis that probability belongs to an interval is a new statistical test, developed in this study.
• Natural video & multimedia noise sources: Modelling is achieved in the 2D DCT (2D Discrete Cosine Transform) and 2D DWT (2D Discrete Wavelet Transform) domains and is expressed as Gaussian mixture whose parameters are estimated by an EM (expectation maximisation) algorithm. The accuracy of the results is proved by relative mean squared errors lower than 5%, computed on both models and the related probabilities. The stationarity/ergodicity proof is objectively assed by: (1) obtaining mean squared errors lower than 15% (these errors were computed on both models and the related probabilities), (2) revealing information based similarity measures close to 0 (the Hellinger distance and the Kullback Leibler divergence) and (3) by a posteriori validating the maximum likelihood criterion involved in the EM algorithm with 95% confidence limit estimation.

A related study follows an application driven approach and is devoted to a clinical driven medical data investigation.

Content enrichment

In its largest acceptation, enriched media refers to all possible associations established between some original data (video, audio, 3D, ...) and some metadata (textual, audio, video, executable codes, ...). Although very attractive from the industrial point of view (iHDTV, computer games, data mining...) the practical deployment of such solutions are mainly restricted by the adaptability, interoperability and backward compatibility constraints.
In order to solve these issues, a theoretical framework is advanced: the in band enrichment, which follows the watermarking philosophy and assumes that the enrichment information is inserted into the very data to be enriched. This new framework not only intrinsically solves the above mentioned constraints but also makes it possible for some theoretical limits to be explored. In this respect, classic information theory tools are adapted/extended in order to compute the capacity for some new types of channels, thus ensuring a theoretical proof of concepts for the in band enrichment.
The fundamental aspects revealed by the studies on natural sources stationarity and content enrichment are further merged in order to bring alive the multimedia content watermarking. The main advances here consist in designing the first method independent with the data peculiarities (uncompressed video/audio/3D) and in the first method robust against transcoding and geometrical attacks processing directly the MPEG 4 AVC (H.264) stream.

Exchange and interaction

Nowadays, a general discrepancy can be noticed between the large variety of mobile thin devices (cell phones, PDA, ...) and the quit poor multimedia experience they offer to the user.
This dead lock has been here addressed from the content point of view. By providing architectural and functional advances, an objective assessment study on the multimedia standard (MPEG 4 BiFS and LASeR) opportunity for deploying such solution has been performed. The overall result can be considered as both a proof of concepts for these emerging technologies and as an enabler for standard evolutions.

---

 

«Indoor Positioning Using GNSS Based Transmitters»

Who ?

The habilitation was defended by Nel Samama, Associate Professor in EPH Department

When ?

Monday 22nd of March, 2010

Where ?

At Telecom ParisTech - 46, rue Barrault - 75013 Paris

Board

• Mr Michel Kasser, Professor, ENSG - Reporter
• Mr Bertrand Merminod, Professor, EPFL - Reporter
• Mr François Peyret, Research Director, Centre de Nantes du LCPC - Reporter
• Mr Aziz Benlarbi-Delaï, Professor, UPMC - Board Member
• Mr Arnaud de la Fortelle, Dean of CAOR, Mines ParisTech - Board Member
• Mr Roger Pagny, CGDD/SDI, Dean of the Satellite Applications Mission- Board Member

Works abstract

Ambitions and Goals

Current localisation systems face the difficulty of achieving the positioning function in various environments with a consistent quality: this probably limits the development of location-based services. Indoor positioning is the most representative case of this difficulty: the diversity of solutions and their maturity show that the "technical road" has not yet been paved. Moreover, the directions being taken by large industrial companies show limitations that are reviving the interest in innovative research. However, the choice of a technical solution is bound to be made within a few years to coincide with the maturity of the Galileo program and the widespread use of mobile internet.

The advantage of using a satellite navigation receiver for locating in areas of non-coverage of GNSS signals is that it is then possible to talk of a "universal" system operating in all environments and using a single technical means. A local infrastructure seems to be needed in order to provide a true expanded coverage: the two types of infrastructure we have studied are repeaters and pseudolites. Repeater-based approaches (GPS or Galileo) are a real alternative to systems like high sensitivity GPS or Assisted-GPS. The latest works on repeaters have led to the following significant achievements:

• the development of the architecture of an indoor positioning system using GNSS signals with new theoretical approaches to signal processing in order to improve the accuracy of the measurement of the pseudo-ranges (works on noise and multipath);
• experimental demonstrations of operation and validation of performance (especially with the help of CNES and specific equipment from the CNES too);
• the development of a mobile demonstrator.

The overall objective of our work is to optimize the detection of the code phase jumps that occur in the repeater-based positioning solution. The three directions that are followed are respectively the analysis of the behaviour of receivers, the simulation and modelling of receiver main loops and the analysis of multipath.

Main results

The main achievements of recent years are as follows:

• Works concerning the difficult problem of multipath. Of course, having a way to cope with multipath indoors is of uppermost importance since this typical environment is highly affected by this disturbing phenomenon. The various works carried out were first oriented towards the full evaluation of the relative importance of multipath in indoor environments. The main topics dealt with were the following: electromagnetic modelling of various indoor environments, simulation of propagation in these environments to assess the importance of multipath, evaluation of deployment rules of repeater antennas in order to reduce multipath, analysis of the effect of "small" displacements (of the order of magnitude of the wavelength of the GPS L1 signal) around a given position of the transponders and the receiver to determine the sensitivity of multiple paths to these movements. The last work consisted in a joint simulation including propagation and receiver new loop behaviour in order to evaluate some new approaches.
• Simulation of a GPS receiver: following the design of a receiver with Matlab/Simulink, we have undertaken to propose new architectures of tracking loops in order to optimize the detection and measurement precision of the typical jumps that occur in the code phase during the transition from one repeater to the next. The three following architectures have been simulated: the open loop, the short multipath insensitive code loop (SMICL), and the loop insensitive to larger multipath (MIDDL). Two publications, accepted by IEEE, and three patent applications have resulted from this work (Nabil Jardak's thesis).
• Implementation of the above mentioned loops and various other calculations on the receiver. The works cited above have naturally given rise to practical implementations on a software receiver, the NordNav R30.
• Real deployments: several experiments were conducted on the premises of Telecom SudParis and at the CNES in Toulouse.
• Works on "Repealites": the main limitation of repeaters in order to improve the positioning accuracy is the impossibility to carry out measurements of carrier phase, the latter being "lost" during the transition from one repeater to the next. The idea of a continuing transmission is so natural. We are working on a time shifted simultaneous transmission to allow, while retaining the benefits of the repeater approach, carrier phase measurements, as for pseudolites. This work is the subject of the thesis of Alexandre Vervisch-Picois, which led to an article in IEEE Signal Processing journal. A patent is being filed on a technique to reduce the near-far effect in a system using pseudolites.

Partnerships

• Strong links with other research institutes (CNES, EPFL, Telecom ParisTech, Fraunhofer, etc.) and manufacturers in the field of localisation (Thales Alenia Space, Septentrio, M3Systems, uBlox, Deveryware, etc.).
• Exchanges underway with international universities: PLAN group at the University of Calgary, Canada and with the SNAP group at the University of New South Wales, Australia