05-Électronique et Télécommunications

Permanent URI for this collectionhttps://dspace.univ-tebessa.dz/handle/123456789/148

Browse

Now showing 1 - 7 of 7

Réalisation d'un fauteuil roulant intelligent (freemove)
(UNIVERSITE DE ECHAHID CHEIKH LARBI TEBESSI, 2026-06) BAIZID SAMAH/ Chaouchi Douaa/encadré par MAAMRI Mahmoud
Les difficultés de mobilité constituent un défi majeur qui altère considérablement la qualité de vie des personnes à mobilité réduite. Les fauteuils roulants traditionnels présentent encore des lacunes en coté de sécurité et des fonctionnalités limitées, réduisant l'autonomie de l'utilisateur et augmentant les risques lors de ses déplacements. Ce projet propose la réalisation et le développement d'un fauteuil roulant intelligent, basé sur des technologies électroniques intégrées et des systèmes de capteurs avancés, visant à améliorer la mobilité des personnes handicapées, notamment des personnes amputées. Ce fauteuil roulant intelligent représente un modèle alliant ingénierie électronique et applications d'assistance médicale ; il est considéré comme une solution technologique innovante susceptible d'être développée en un produit industriel destiné au marché local.
Berkani lamis BOUKHRIS HAIFA Aoulmi Chahinda / encadré par CHERIET Leyla
(2026-06) Berkani lamis/ BOUKHRIS HAIFA /encadré par Aoulmi Chahinda
À l’ère de l’intelligence artificielle générative, la prolifération des manipulations d’images numériques, notamment les Deepfakes, pose un défi majeur pour la criminalistique informatique (Digital Forensics). Ce mémoire de Master présente une approche hybride robuste dédiée à la détection automatique de falsifications d’images, en combinant l’analyse spatio-fréquentielle et l’apprentissage profond. Notre contribution fondamentale repose sur l’intégration de la Transformée en Ondelettes Discrète (DWT) en amont des Réseaux de Neurones Convolutifs (CNN). En écartant l’approximation globale basse fréquence pour forcer le réseau à focaliser son attention sur les sous-bandes de détails fréquentielles (LH,HL,HH), le système parvient à isoler les micro-anomalies géométriques et statistiques laissées par les algorithmes de manipulation. Les simulations rigoureuses menées sous l’environnement MATLAB sur la base de données de référence Celeb- DF v2 valident l’efficacité de cette fusion. En utilisant l’ondelette de Daubechies (db4), le modèle basé sur l’apprentissage résiduel ResNet-18 surclasse les architectures standard et SqueezeNet, en atteignant une précision remarquable de 95,42% (jusqu’à 96,15%), avec un score AUC de 0,978 et un taux de fausses alarmes (FPR) réduit à seulement 4,00%.
Optimisation d’un système biométrique sécurisé à l’aide de filtres orientés
(UNIVERSITE DE ECHAHID CHEIKH LARBI TEBESSI, 2026-06-09) MAALLEM SAFOUA /Si Said Ania/ encadré par Abdallah Meraoumia
Les systèmes biométriques occupent aujourd’hui une place importante dans les domaines de la sécurité et de l’authentification grâce à leur capacité à identifier les individus à partir de caractéristiques physiologiques ou comportementales. Cependant, ces systèmes restent exposés à plusieurs menaces telles que le spoofing, le vol des templates biométriques et les attaques sur les canaux de transmission. Ce mémoire présente une étude des systèmes biométriques, de leurs vulnérabilités ainsi que des principales techniques de protection utilisées pour renforcer leur sécurité. Une attention particulière est accordée à la protection des templates biométriques à travers des mécanismes de chiffrement et des méthodes anti-spoofing. Dans ce contexte, une approche sécurisée est proposée en utilisant les projections spatiales guidées par le chaos, l’optimisation par essaim de particules (PSO) ainsi qu’un schéma de chiffrement adapté à la protection des caractéristiques biométriques. Cette approche vise à améliorer la sécurité, l’intégrité et la robustesse du système face aux différentes attaques
Chaotic Encryption for Secure Image Transmission in Free-Space Optical Systems
(UNIVERSITE DE ECHAHID CHEIKH LARBI TEBESSI, 2026-06-09) Chettouh Chayma ,Harbaoui Nesrine / encadré par Hanane Djellab
This work presents a secure Free-Space Optical (FSO) communication system based on a four-dimensional memristor-based hyperchaotic image encryption scheme. The proposed framework combines hyperchaotic cryptography with optical wireless communication to ensure both high-level data security and reliable image transmission under atmospheric disturbances. A 4D hyperchaotic system was designed and analyzed using time-series analysis, phase-space trajectories, and Lyapunov exponents, confirming strong randomness and high sensitivity to initial conditions. Three image encryption methods based on XOR diffusion, arithmetic addition, and QR matrix decomposition were developed and evaluated using entropy, correlation, NPCR, UACI, and key sensitivity analyses. The obtained entropy values reached 7.9987, 7.9983, and 7.9893, while NPCR values exceeded 99.55% and UACI values remained close to 33%, demonstrating excellent security performance and strong resistance against statistical and differential attacks. In addition, the proposed system provided a large key space estimated at approximately ( 10112.). The proposed encryption framework was integrated into an FSO communication system using MATLAB and OptiSystem environments. The obtained results showed that clear weather achieved the best transmission performance, maintaining BER values below ( 10−5.) and Q-factor values above 4 up to approximately 1.6 km, with a BER of (8.68 × 10−22.) and a Q-factor of 9.51 at 1 km. Under moderate rain conditions, the acceptable communication range decreased to approximately 1.1 km with a BER of (4.27 × 10−9.) and a Q-factor of 5.75, while snow conditions reduced the effective range to approximately 0.4 km with a BER of (1.89× 10−4.) and a Q-factor of 3.56 due to severe atmospheric attenuation. Overall, the obtained results demonstrate that the proposed QR-based hyperchaotic encryption scheme provides an effective compromise between cryptographic security, diffusion capability, and transmission robustness, making the proposed framework a promising solution for next-generation secure optical wireless communication systems.
Miniaturisation des Antennes Patch à l'Aide des Métamatériaux pour les Applications Sans Fil
(UNIVERSITE DE ECHAHID CHEIKH LARBI TEBESSI, 2026-06-09) Nabet Chourouk / Khediri Ramzi / encadré par FERROUDJI Karim
This final year project focuses on the design, optimization, and miniaturization of a high-performance planar patch antenna for modern communication applications (5G, WiMAX, 4G/LTE, IoT, …). The main objective of this research was to overcome the major theoretical limitation that systematically links reducing the physical size of antennas to a degradation in their performance. To meet this challenge, we exploited the novel electromagnetic properties of metamaterials. The scientific approach adopted consisted of first designing a conventional reference antenna optimized to resonate at an initial frequency of 5 GHz. Subsequently, a complementary split-ring resonator (CSRR) was precisely integrated into the ground plane. The electromagnetic coupling thus generated made it possible to confine the energy and successfully shift the fundamental frequency to 2.50 GHz, thereby achieving a remarkable miniaturization rate without compromising the overall radiative integrity of the structure. Numerical simulations performed using ANSYS HFSS software confirm the robustness and technical viability of the proposed architecture, demonstrating that it constitutes an efficient
Evaluating Explainable Artificial Intelligence (XAI) in the Network Security Domain A Case Study
(UNIVERSITE DE ECHAHID CHEIKH LARBI TEBESSI, 2026-06-09) Haouichi Salah Eddine / encadré par Aouiche Chaima
The rapid expansion of Internet of Things infrastructure has fundamentally altered the network security threat landscape, creating environments in which the volume, velocity, and diversity of malicious traffic exceed the capacity of rule-based detection systems and human analyst teams alike. Machine learning offers a powerful alternative, but the models that achieve the highest detection rates — deep neural networks and large ensemble architectures — are precisely those whose internal reasoning is most opaque, creating a critical gap between predictive capability and operational interpretability that undermines the trust and accountability requirements of real security operations. This thesis investigates whether explainable artificial intelligence can bridge that gap in an IoT network intrusion detection context. A comparative experimental framework was constructed in which three architecturally distinct models — a Deep Neural Network, a Bagging ensemble based on Random Forest, and a two-level Stacking ensemble — were trained and evaluated on the CIC IoT-DIAD 2024 dataset, comprising 27,713,532 labelled network flow records described by 36 traffic-derived features. Each model was assessed not only on standard detection metrics but on the quality, consistency, and operational interpretability of the SHAP attributions it produced, using architecture-appropriate explainer classes: TreeExplainer for Bagging, DeepExplainer for DNN, and PermutationExplainer for Stacking. All three architectures exceeded 97% accuracy and 97% F1-score on the balanced held-out test set. The Bagging ensemble achieved the highest precision (99.76%) and the lowest false positive count (188), while the Stacking ensemble achieved the highest recall (96.71%) and the lowest false negative count (2,622). SHAP analysis identified three features — SYN Flag Count, Bwd Packet Length Max, and Flow IAT Min — that appeared consistently in the top 10 rankings of all three architectures, providing a cross-model consensus that is both statistically robust and operationally interpretable in terms of known IoT attack mechanisms. Spearman rank correlation analysis revealed that attribution consistency is architecture-dependent: the tree-based Bagging and Stacking models produce strongly correlated rankings (ρ = 0.7521), while the DNN's attribution profile diverges significantly from both ensembles (ρ ≈ 0.12–0.20). The Bagging ensemble is identified as the recommended architecture for the evaluated scenario, optimising across detection reliability, explanation exactness, computational feasibility, and cross-model attribution consistency simultaneously. The study demonstrates that XAI-enhanced intrusion detection is both technically viable and operationally actionable for IoT security environments, and introduces cross-model rank correlation analysis as a practical framework for assessing explanation robustness in multi-model security deployments.
Enhanced medical image diagnosis using a deep learning approach based on autoencoders
(UNIVERSITE DE ECHAHID CHEIKH LARBI TEBESSI, 2026-06-09) Ghalloussi Wissal / Filali Hiba encadré par Haouam Lotfi
Medical image analysis is essential for disease diagnosis, but its interpretation remains challenging. This thesis investigates the use of deep learning, particularly Convolutional Neural Networks (CNNs) and autoencoders, to automate medical image classification. Two case studies are presented: brain tumor classification from MRI images and pneumonia detection from chest X-ray images. The obtained results demonstrate high classification accuracy and confirm the potential of these approaches for developing computer-aided diagnosis systems.