{"id":7730,"date":"2026-11-19T08:04:31","date_gmt":"2026-11-19T07:04:31","guid":{"rendered":"https:\/\/samovar.telecom-sudparis.eu\/?p=7730"},"modified":"2026-11-19T08:04:32","modified_gmt":"2026-11-19T07:04:32","slug":"avis-de-soutenance-de-madame-reem-nabha","status":"publish","type":"post","link":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/2026\/11\/19\/avis-de-soutenance-de-madame-reem-nabha\/","title":{"rendered":"AVIS DE SOUTENANCE de Madame Reem NABHA"},"content":{"rendered":"\n<h2 class=\"wp-block-heading\">L&rsquo;Ecole doctorale : Ecole Doctorale de l&rsquo;Institut Polytechnique de Paris<br><br>et le Laboratoire de recherche SAMOVAR &#8211; Services r\u00e9partis, Architectures, Mod\u00e9lisation, Validation, Administration des R\u00e9seaux<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">pr\u00e9sentent<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">l\u2019AVIS DE SOUTENANCE de Madame Reem NABHA<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Autoris\u00e9e \u00e0 pr\u00e9senter ses travaux en vue de l\u2019obtention du Doctorat de l&rsquo;Institut Polytechnique de Paris, pr\u00e9par\u00e9 \u00e0 l&rsquo;Institut Polytechnique de Paris T\u00e9l\u00e9com SudParis en :<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">R\u00e9seaux, Informations et Communications<\/h2>\n\n\n\n<h1 class=\"wp-block-heading\">\u00ab Pr\u00e9servation de la vie priv\u00e9e \u00e0 l\u2019aide de l\u2019apprentissage f\u00e9d\u00e9r\u00e9 d\u00e9centralis\u00e9 pour les syst\u00e8mes de sant\u00e9 \u00bb<\/h1>\n\n\n\n<p class=\"wp-block-paragraph\">le VENDREDI 20 NOVEMBRE 2026 \u00e0 9h30<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u00e0<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Auditorium, Doctoral School of Science &amp; Technology, Lebanese University<br>Lebanese University, Hadath Campus, Lebanon<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/teams.microsoft.com\/meet\/3852974113189337?p=U7My2TNcnkPFfSrKbU\">https:\/\/teams.microsoft.com\/meet\/3852974113189337?p=U7My2TNcnkPFfSrKbU<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Meeting ID: <a href=\"callto:385 297 411 318 933 7\">385 297 411 318 933 7<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Passcode: iz2qk6SP<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Membres du jury :<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>M. Anis&nbsp;LAOUITI<\/strong>, Professeur, T\u00e9l\u00e9com SudParis \u2013 SAMOVAR, Institut Polytechnique de Paris, FRANCE &#8211; Directeur de th\u00e8se<br><strong>Mme Tara&nbsp;ALI-YAHIYA<\/strong>, Ma\u00eetresse de conf\u00e9rences, Universit\u00e9 Paris-Saclay, FRANCE &#8211; Rapporteur<br><strong>M. Abbass&nbsp;NASSER<\/strong>, Ma\u00eetre de conf\u00e9rences, Holy Spirit University of Kaslik (USEK), Lebanon, LIBAN &#8211; Rapporteur<br><strong>M. Mohammad&nbsp;BEYROUTI<\/strong>, Enseignant-Chercheur, Universit\u00e9 de Technologie de Troyes (UTT) \u2013 Laboratoire LIST3N, FRANCE &#8211; Examinateur<br><strong>M. Riadh&nbsp;DHAOU<\/strong>, Professeur des universit\u00e9s, Toulouse INP \u2013 ENSEEIHT \/ IRIT, France, FRANCE &#8211; Examinateur<br><strong>Mme Zainab&nbsp;ASSAGHIR<\/strong>, Professeur, Lebanese University \u2013 Faculty of Science, LIBAN &#8211; Examinateur<br><strong>M. Abed Ellatif&nbsp;SAMHAT<\/strong>, Professeur, Lebanese University \u2013 Faculty of Engineering, LIBAN &#8211; Directeur de th\u00e8se<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">\u00ab Pr\u00e9servation de la vie priv\u00e9e \u00e0 l\u2019aide de l\u2019apprentissage f\u00e9d\u00e9r\u00e9 d\u00e9centralis\u00e9 pour les syst\u00e8mes de sant\u00e9 \u00bb<\/h2>\n\n\n\n<h2 class=\"wp-block-heading\">pr\u00e9sent\u00e9 par Madame Reem NABHA<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>R\u00e9sum\u00e9 :<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Les syst\u00e8mes de sant\u00e9 num\u00e9riques s\u2019appuient de plus en plus sur des donn\u00e9es distribu\u00e9es et sur l\u2019apprentissage automatique afin de soutenir le diagnostic et la prise de d\u00e9cision, tandis que la sensibilit\u00e9 des informations m\u00e9dicales fait de la pr\u00e9servation de la confidentialit\u00e9 un enjeu majeur. L\u2019apprentissage f\u00e9d\u00e9r\u00e9 (Federated Learning, FL) permet l\u2019entra\u00eenement collaboratif de mod\u00e8les sans \u00e9change de donn\u00e9es brutes, mais les approches f\u00e9d\u00e9r\u00e9es centralis\u00e9es conventionnelles d\u00e9pendent encore d\u2019un serveur central d\u2019agr\u00e9gation et n\u00e9cessitent souvent l\u2019\u00e9change d\u2019une quantit\u00e9 importante d\u2019informations sur le mod\u00e8le, ce qui soul\u00e8ve des d\u00e9fis en mati\u00e8re de confidentialit\u00e9, de s\u00e9curit\u00e9 et de communication. Cette th\u00e8se propose un cadre d\u2019apprentissage f\u00e9d\u00e9r\u00e9 d\u00e9centralis\u00e9 avec partage partiel de la t\u00eate de classification (Decentralized Federated Learning with Partial Head Sharing, DFL-PHS), con\u00e7u pour l\u2019apprentissage collaboratif dans des environnements sensibles \u00e0 la confidentialit\u00e9. Au lieu de partager l\u2019ensemble des param\u00e8tres du mod\u00e8le, les participants n\u2019\u00e9changent qu\u2019une proportion configurable des poids des derni\u00e8res couches denses, tandis que les donn\u00e9es brutes, les couches d\u2019extraction de caract\u00e9ristiques et les param\u00e8tres non partag\u00e9s restent locaux. Chaque participant effectue l\u2019entra\u00eenement local, \u00e9change directement des mises \u00e0 jour partielles avec ses pairs et r\u00e9alise localement une agr\u00e9gation pond\u00e9r\u00e9e en fonction de la taille des ensembles de donn\u00e9es, sans recourir \u00e0 un serveur central. Le cadre est principalement \u00e9valu\u00e9 sur le jeu de donn\u00e9es COVID-19 Radiography \u00e0 l\u2019aide d\u2019un mod\u00e8le bas\u00e9 sur MobileNetV2. L\u2019apprentissage local, l\u2019apprentissage f\u00e9d\u00e9r\u00e9 centralis\u00e9, l\u2019entra\u00eenement sur un site unique et DFL-PHS sont compar\u00e9s selon diff\u00e9rentes tailles de jeux de donn\u00e9es et des taux de partage de param\u00e8tres de 25 %, 50 %, 75 % et 100 %. Les exp\u00e9riences \u00e9valuent les performances pr\u00e9dictives ainsi que la quantit\u00e9 d\u2019informations du mod\u00e8le \u00e9chang\u00e9es. La confidentialit\u00e9 et la s\u00e9curit\u00e9 sont \u00e9galement \u00e9tudi\u00e9es \u00e0 travers des attaques d\u2019inf\u00e9rence d\u2019appartenance, d\u2019inversion de gradient, d\u2019inversion de poids, de retournement d\u2019\u00e9tiquettes et d\u2019empoisonnement des poids, afin d\u2019analyser le comportement de DFL-PHS face \u00e0 des menaces d\u2019inf\u00e9rence et d\u2019int\u00e9grit\u00e9. Afin d\u2019examiner si le principe sous-jacent de partage partiel reste applicable au-del\u00e0 de l\u2019imagerie m\u00e9dicale, le cadre est \u00e9galement \u00e9valu\u00e9 sur le jeu de donn\u00e9es UNSW-NB15 de d\u00e9tection d\u2019intrusions r\u00e9seau, \u00e0 l\u2019aide d\u2019un mod\u00e8le de repr\u00e9sentation tabulaire et d\u2019une t\u00eate de classification dense. Cette exp\u00e9rimentation inter-domaines examine le cadre avec une modalit\u00e9 de donn\u00e9es diff\u00e9rente, une architecture de mod\u00e8le diff\u00e9rente et une distribution h\u00e9t\u00e9rog\u00e8ne des donn\u00e9es entre les clients. Dans l\u2019ensemble, cette th\u00e8se montre qu\u2019une collaboration d\u00e9centralis\u00e9e avec partage partiel de la t\u00eate de classification peut maintenir un apprentissage collaboratif efficace tout en r\u00e9duisant la d\u00e9pendance \u00e0 un serveur central et en limitant la quantit\u00e9 d\u2019informations du mod\u00e8le expos\u00e9es lors des communications. Les r\u00e9sultats soulignent \u00e9galement que le partage partiel ne constitue pas une garantie formelle de s\u00e9curit\u00e9 et que la confidentialit\u00e9 ainsi que la robustesse doivent \u00eatre \u00e9valu\u00e9es explicitement dans des conditions adversariales r\u00e9alistes.<br><strong>Abstract :<\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Digital healthcare systems increasingly rely on distributed data and machine learning to support diagnosis and decision-making, while the sensitivity of medical information makes privacy preservation a major concern. Federated Learning (FL) enables collaborative model training without exchanging raw data, but conventional centralized FL still depends on a central aggregation server and often requires substantial model-information exchange, creating privacy, security, and communication challenges. This thesis proposes Decentralized Federated Learning with Partial Head Sharing (DFL-PHS), a peer-to-peer framework designed for privacy-sensitive collaborative learning. Instead of sharing complete model parameters, participants exchange only a configurable proportion of the weights from the final dense layers, while raw data, feature-extraction layers, and non-shared parameters remain local. Each participant performs local training, exchanges partial updates directly with its peers, and carries out dataset-size weighted aggregation locally without relying on a central server. The framework is primarily evaluated on the COVID-19 Radiography Dataset using a MobileNetV2-based model. Local Learning, Centralized Federated Learning, Single-Site Training, and DFL-PHS are compared across different dataset sizes and parameter-sharing ratios of 25%, 50%, 75%, and 100%. The experiments assess predictive performance together with the amount of model information exchanged. Privacy and security are further investigated through membership inference, gradient inversion, weight inversion, label flipping, and weight poisoning attacks, allowing the behavior of DFL-PHS to be examined under both inference and integrity threats. To investigate whether the underlying partial-sharing principle remains applicable beyond medical imaging, the framework is also evaluated on the UNSW-NB15 network intrusion detection dataset using a tabular representation model and a dense classification head. This cross-domain experiment examines the framework under a different data modality, model architecture, and heterogeneous client distribution. Overall, the thesis shows that decentralized collaboration with partial head sharing can maintain effective collaborative learning while reducing central-server dependency and limiting the amount of model information exposed during communication. The results also emphasize that partial sharing is not a formal security guarantee and that privacy and robustness must be evaluated explicitly under realistic adversarial conditions.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>L&rsquo;Ecole doctorale : Ecole Doctorale de l&rsquo;Institut Polytechnique de Paris et le Laboratoire de recherche SAMOVAR &#8211; Services r\u00e9partis, Architectures, Mod\u00e9lisation, Validation, Administration des R\u00e9seaux pr\u00e9sentent l\u2019AVIS DE SOUTENANCE de Madame Reem NABHA Autoris\u00e9e \u00e0 pr\u00e9senter ses travaux en vue de l\u2019obtention du Doctorat de l&rsquo;Institut Polytechnique de Paris, pr\u00e9par\u00e9 \u00e0 l&rsquo;Institut Polytechnique de Paris [&hellip;]<\/p>\n","protected":false},"author":4,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ocean_front_end_style_editor":"no","ocean_post_layout":"","ocean_both_sidebars_style":"","ocean_both_sidebars_content_width":0,"ocean_both_sidebars_sidebars_width":0,"ocean_sidebar":"","ocean_second_sidebar":"","ocean_disable_margins":"enable","ocean_add_body_class":"","ocean_shortcode_before_top_bar":"","ocean_shortcode_after_top_bar":"","ocean_shortcode_before_header":"","ocean_shortcode_after_header":"","ocean_has_shortcode":"","ocean_shortcode_after_title":"","ocean_shortcode_before_footer_widgets":"","ocean_shortcode_after_footer_widgets":"","ocean_shortcode_before_footer_bottom":"","ocean_shortcode_after_footer_bottom":"","ocean_display_top_bar":"default","ocean_display_header":"default","ocean_header_style":"","ocean_center_header_left_menu":"","ocean_custom_header_template":"","ocean_custom_logo":0,"ocean_custom_retina_logo":0,"ocean_custom_logo_max_width":0,"ocean_custom_logo_tablet_max_width":0,"ocean_custom_logo_mobile_max_width":0,"ocean_custom_logo_max_height":0,"ocean_custom_logo_tablet_max_height":0,"ocean_custom_logo_mobile_max_height":0,"ocean_header_custom_menu":"","ocean_menu_typo_font_family":"","ocean_menu_typo_font_subset":"","ocean_menu_typo_font_size":0,"ocean_menu_typo_font_size_tablet":0,"ocean_menu_typo_font_size_mobile":0,"ocean_menu_typo_font_size_unit":"px","ocean_menu_typo_font_weight":"","ocean_menu_typo_font_weight_tablet":"","ocean_menu_typo_font_weight_mobile":"","ocean_menu_typo_transform":"","ocean_menu_typo_transform_tablet":"","ocean_menu_typo_transform_mobile":"","ocean_menu_typo_line_height":0,"ocean_menu_typo_line_height_tablet":0,"ocean_menu_typo_line_height_mobile":0,"ocean_menu_typo_line_height_unit":"","ocean_menu_typo_spacing":0,"ocean_menu_typo_spacing_tablet":0,"ocean_menu_typo_spacing_mobile":0,"ocean_menu_typo_spacing_unit":"","ocean_menu_link_color":"","ocean_menu_link_color_hover":"","ocean_menu_link_color_active":"","ocean_menu_link_background":"","ocean_menu_link_hover_background":"","ocean_menu_link_active_background":"","ocean_menu_social_links_bg":"","ocean_menu_social_hover_links_bg":"","ocean_menu_social_links_color":"","ocean_menu_social_hover_links_color":"","ocean_disable_title":"default","ocean_disable_heading":"default","ocean_post_title":"","ocean_post_subheading":"","ocean_post_title_style":"","ocean_post_title_background_color":"","ocean_post_title_background":0,"ocean_post_title_bg_image_position":"","ocean_post_title_bg_image_attachment":"","ocean_post_title_bg_image_repeat":"","ocean_post_title_bg_image_size":"","ocean_post_title_height":0,"ocean_post_title_bg_overlay":0.5,"ocean_post_title_bg_overlay_color":"","ocean_disable_breadcrumbs":"default","ocean_breadcrumbs_color":"","ocean_breadcrumbs_separator_color":"","ocean_breadcrumbs_links_color":"","ocean_breadcrumbs_links_hover_color":"","ocean_display_footer_widgets":"default","ocean_display_footer_bottom":"default","ocean_custom_footer_template":"","ocean_post_oembed":"","ocean_post_self_hosted_media":"","ocean_post_video_embed":"","ocean_link_format":"","ocean_link_format_target":"self","ocean_quote_format":"","ocean_quote_format_link":"post","ocean_gallery_link_images":"on","ocean_gallery_id":[],"footnotes":""},"categories":[286,402],"tags":[],"class_list":["post-7730","post","type-post","status-publish","format-standard","hentry","category-fractualites-ennews-fr","category-seminaires-ness-2013-fr","entry"],"_links":{"self":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/7730","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/users\/4"}],"replies":[{"embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/comments?post=7730"}],"version-history":[{"count":1,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/7730\/revisions"}],"predecessor-version":[{"id":7731,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/posts\/7730\/revisions\/7731"}],"wp:attachment":[{"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/media?parent=7730"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/categories?post=7730"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/samovar.telecom-sudparis.eu\/index.php\/wp-json\/wp\/v2\/tags?post=7730"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}