Artificial intelligence-based system for accident prevention on railways / Hanane Hamar (2020) 

Public ISBD Titre : Artificial intelligence-based system for accident prevention on railways Type de document : document électronique Auteurs : Hanane Hamar, Auteur ; Sarah Si Youcef, Auteur ; Sid-Ahmed Berrani, Directeur de thèse Editeur : [S.l.] : [s.n.] Année de publication : 2020 Importance : 1 fichier PDF (13 M) Présentation : ill. Note générale : Mode d'accès : accès au texte intégral par intranet. Mémoire de Projet de Fin d’Études : Électronique : Alger, École Nationale Polytechnique : 2020 Bibliogr. f. 66 - 70 Langues : Anglais (eng) Mots-clés : Train Railway Computer Vision Machine learning Accident prevention Object Detection Evaluation Distance Estimation Index. décimale : PN00620 Résumé : Railway transport suffers from a major problem which is accidents, where the losses in hu-man lives and the materials are enormous. To solve this problem or at least reduce the number of accidents, railway safety has proposednumerous initiatives with different methods and technologies. In this work, a method usingartificial intelligence is proposed. The system developed must comply with numerous conditions contrary to the systems de-signed for the autonomous vehicle sector. One of the key challenges is long-range obstacledetection. Sensor technology in current land transport research is able to look 200 m ahead. A system that combines two types of cameras; RGB and thermal; is suggested to consider thedifferent requirements. these ones consist of the different illumination and weather conditions,and also the range needed to detect obstacle on time. This work is divided into two parts. The first one is the obstacle detection. To do so,three common object detection algorithms have been evaluated and compared (effectivenessand efficiency comparison). The algorithm with the lowest FAR( false Alarm rate), the bestDR (Detection Rate), and that achieves the shortest execution time has been chosen. Theobjective is to build a reliable system that can operate in real-time.The second part focuses on the estimation of the distance between objects and the train usingthe DisNet algorithm. Here, two evaluation methods have been used in order to assess theprecision of the estimation. Artificial intelligence-based system for accident prevention on railways [document électronique] / Hanane Hamar, Auteur ; Sarah Si Youcef, Auteur ; Sid-Ahmed Berrani, Directeur de thèse . - [S.l.] : [s.n.], 2020 . - 1 fichier PDF (13 M) : ill. Mode d'accès : accès au texte intégral par intranet. Mémoire de Projet de Fin d’Études : Électronique : Alger, École Nationale Polytechnique : 2020 Bibliogr. f. 66 - 70 Langues : Anglais (eng) Mots-clés : Train Railway Computer Vision Machine learning Accident prevention Object Detection Evaluation Distance Estimation Index. décimale : PN00620 Résumé : Railway transport suffers from a major problem which is accidents, where the losses in hu-man lives and the materials are enormous. To solve this problem or at least reduce the number of accidents, railway safety has proposednumerous initiatives with different methods and technologies. In this work, a method usingartificial intelligence is proposed. The system developed must comply with numerous conditions contrary to the systems de-signed for the autonomous vehicle sector. One of the key challenges is long-range obstacledetection. Sensor technology in current land transport research is able to look 200 m ahead. A system that combines two types of cameras; RGB and thermal; is suggested to consider thedifferent requirements. these ones consist of the different illumination and weather conditions,and also the range needed to detect obstacle on time. This work is divided into two parts. The first one is the obstacle detection. To do so,three common object detection algorithms have been evaluated and compared (effectivenessand efficiency comparison). The algorithm with the lowest FAR( false Alarm rate), the bestDR (Detection Rate), and that achieves the shortest execution time has been chosen. Theobjective is to build a reliable system that can operate in real-time.The second part focuses on the estimation of the distance between objects and the train usingthe DisNet algorithm. Here, two evaluation methods have been used in order to assess theprecision of the estimation.

Réservation

Réserver ce document

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité	Spécialité	Etat_Exemplaire
EP00076	PN00620	Ressources électroniques	Bibliothèque centrale	Projet Fin d'Etudes	Disponible	Electronique	Téléchargeable

Documents numériques

HAMAR.Hanane_SI-YOUCEF.Sarah.pdf URL

Assessment of deepfake detection techniques / Abderezak Mechenet (2024) 

Public ISBD Titre : Assessment of deepfake detection techniques : a study of performance and generalisation Type de document : document électronique Auteurs : Abderezak Mechenet, Auteur ; Sid-Ahmed Berrani, Directeur de thèse Editeur : [S.l.] : [s.n.] Année de publication : 2024 Importance : 1 fichier PDF (9 Mo) Présentation : ill. Note générale : Mode d'accès : accès au texte intégral par intranet. Mémoire de Projet de Fin d’Études : Electronique : Alger, École Nationale Polytechnique : 2024 Bibliogr. p. 124 - 131 . - Annexe p. 132 - 134 Langues : Anglais (eng) Mots-clés : Deepfake detection Generalisation Media content Index. décimale : PN00224 Résumé : Deepfakes are the fruit of development in the fields of Artificial Intelligence and Deep Learning, their appearance created a new field: deepfake detection, a domain that specialises in the verification of the authenticity of media content. Convolutional Neural Networks based detectors aim at detecting artifacts left by the generation process to draw conclusions on the multimedia. Our study was centralised around the assessment of existing deepfake detectors, aiming to evaluate and improve our diagnosed problem which is the generalisation ability across multiple datasets, our final product was put to the test on external videos to conclude on the hypothesis established. Assessment of deepfake detection techniques : a study of performance and generalisation [document électronique] / Abderezak Mechenet, Auteur ; Sid-Ahmed Berrani, Directeur de thèse . - [S.l.] : [s.n.], 2024 . - 1 fichier PDF (9 Mo) : ill. Mode d'accès : accès au texte intégral par intranet. Mémoire de Projet de Fin d’Études : Electronique : Alger, École Nationale Polytechnique : 2024 Bibliogr. p. 124 - 131 . - Annexe p. 132 - 134 Langues : Anglais (eng) Mots-clés : Deepfake detection Generalisation Media content Index. décimale : PN00224 Résumé : Deepfakes are the fruit of development in the fields of Artificial Intelligence and Deep Learning, their appearance created a new field: deepfake detection, a domain that specialises in the verification of the authenticity of media content. Convolutional Neural Networks based detectors aim at detecting artifacts left by the generation process to draw conclusions on the multimedia. Our study was centralised around the assessment of existing deepfake detectors, aiming to evaluate and improve our diagnosed problem which is the generalisation ability across multiple datasets, our final product was put to the test on external videos to conclude on the hypothesis established.

Réservation

Réserver ce document

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité	Spécialité	Etat_Exemplaire
EP00744	PN00224	Ressources électroniques	Bibliothèque centrale	Projet Fin d'Etudes	Disponible	Electronique	Téléchargeable

Documents numériques

MECHENET.Abderezak.pdf URL

Automation in cybersecurity / Chaimaa Abi (2023) 

Public ISBD Titre : Automation in cybersecurity : deep learning-based approaches for malware family identification Type de document : document électronique Auteurs : Chaimaa Abi, Auteur ; Sid-Ahmed Berrani, Directeur de thèse ; Abdelouahab Boudjellal, Directeur de thèse Editeur : [S.l.] : [s.n.] Année de publication : 2023 Importance : 1 fichier PDF (3.9 Mo) Présentation : ill. Note générale : Mode d'accès : accès au texte intégral par intranet Mémoire de Projet de Fin d’Etudes : Génie Industriel. Data Science-Intelligence Artificielle : Alger, Ecole Nationale Polytechnique : 2023 Bibliogr. P. 65 - 68 Langues : Anglais (eng) Mots-clés : Deep learning Feature extraction Malware analysis Malware classification Malware visualization Multimodal Index. décimale : PI02623 Résumé : The rapid proliferation of malware presents a significant threat to computer systems and data security. The ability to detect and accurately classify malware is crucial for mitigating cyber threats and preventing potential damages. However, traditional methods for malware classification and analysis have shown some limitations in keeping pace with the with the ever-changing landscape of malware. In this thesis, we propose a novel approach that harnesses the power of machine and deep learning techniques for efficient malware classification and offers real-time and automated data-driven solution, enabling proactive measures to efficiently prevent and mitigate cyber threats. Automation in cybersecurity : deep learning-based approaches for malware family identification [document électronique] / Chaimaa Abi, Auteur ; Sid-Ahmed Berrani, Directeur de thèse ; Abdelouahab Boudjellal, Directeur de thèse . - [S.l.] : [s.n.], 2023 . - 1 fichier PDF (3.9 Mo) : ill. Mode d'accès : accès au texte intégral par intranet Mémoire de Projet de Fin d’Etudes : Génie Industriel. Data Science-Intelligence Artificielle : Alger, Ecole Nationale Polytechnique : 2023 Bibliogr. P. 65 - 68 Langues : Anglais (eng) Mots-clés : Deep learning Feature extraction Malware analysis Malware classification Malware visualization Multimodal Index. décimale : PI02623 Résumé : The rapid proliferation of malware presents a significant threat to computer systems and data security. The ability to detect and accurately classify malware is crucial for mitigating cyber threats and preventing potential damages. However, traditional methods for malware classification and analysis have shown some limitations in keeping pace with the with the ever-changing landscape of malware. In this thesis, we propose a novel approach that harnesses the power of machine and deep learning techniques for efficient malware classification and offers real-time and automated data-driven solution, enabling proactive measures to efficiently prevent and mitigate cyber threats.

Réservation

Réserver ce document

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité	Spécialité	Etat_Exemplaire
EP00640	PI02623	Ressources électroniques	Bibliothèque centrale	Projet Fin d'Etudes	Disponible	Data sciences_Intelligence artificielle	Téléchargeable

Documents numériques

ABI.chaimaa.pdf URL

Behavioral biometrics for continuous authentication of smartphone users / Youcef Ouadjer (2026) 

Public ISBD Titre : Behavioral biometrics for continuous authentication of smartphone users Type de document : document électronique Auteurs : Youcef Ouadjer, Auteur ; Mourad Adnane, Directeur de thèse ; Sid-Ahmed Berrani, Directeur de thèse Editeur : [S.l.] : [s.n.] Année de publication : 2026 Importance : 1 fichier PDF (10.4 Mo) Note générale : Mode d'accès : accès au texte intégral par intranet. Thèse de Doctorat : Électronique. Instrumentation : Alger, Ecole Nationale Polytechnique : 2026 Bibliogr. p. 97-108 Thèse confidentielle 2 ans jusqu'à Janvier 2028 Langues : Anglais (eng) Mots-clés : Behavioral biometrics Continuous authentication Multi-modal dataset Self-supervised learning Index. décimale : D000126 Résumé : Smartphone devices have become essential for managing sensitive operations such as on-line banking, accessing medical records, making digital payments, and using government services. As a result, consumers have raised the demand for robust and user-friendly authentication systems. However, traditional smartphone authentication methods such as knowledge-based (e.g., PIN codes, passwords) and static biometric systems (e.g., finger-print, facial recognition) suffer from significant limitations. These methods are vulnerable to smudge and spoofing attacks, and fails to provide ongoing security once a device is un-locked. To address these challenges, we propose the design and evaluation of efficient, multi-modal continuous authentication systems leveraging behavioral biometrics. As a starting point, the research presents a comprehensive review of state-of-the-art meth ods for continuous authentication, highlighting recent progress in behavioral biometric modalities such as hand movement and touch gestures, and identifying limitations in existing datasets. Particularly the lack of synchronized multi-modal behavioral biometric modalities, combined with static biometric characteristics such as facial features. Following, an efficient continuous authentication system is designed, by investigating advanced feature selection to identify most relevant features, showing consistent improvement with the subset of selected features. Further a new continuous authentication system is introduced using self-supervised contrastive learning. The system employs a lightweight convolutional neural network architecture based on depthwise separable convolutions, achieving high accuracy verification and identification tasks while maintaining computational efficiency. By extending the original self-supervised contrastive learning framework introduced in the previous contribution, a multi-modal fusion framework is designed, by combining hand movement and touch gesture data. Fusion is performed at the feature-level, demonstrating robust performance even on small annotated datasets where labeled biometric data is scarce. Lastly, a novel multi-modal dataset, MM-BioSync, is introduced to address the lack of synchronized behavioral biometric public datasets. The dataset integrates data from frontfacing cameras, motion sensors, and touchscreen interactions. Experiments conducted on the dataset reveal that integrating all modalities captured during reading and writing tasks yields the best performance in user verification, underscoring the value of multimodal approaches for continuous authentication. The findings of this thesis highlight the potential of behavioral biometrics and multimodal fusion in enabling continuous user authentication. By addressing key challenges related to model performance, computational efficiency, and dataset availability, this work advances the state-of-the-art and provides a foundation for developing secure and user-friendly authentication solutions on smartphones. Behavioral biometrics for continuous authentication of smartphone users [document électronique] / Youcef Ouadjer, Auteur ; Mourad Adnane, Directeur de thèse ; Sid-Ahmed Berrani, Directeur de thèse . - [S.l.] : [s.n.], 2026 . - 1 fichier PDF (10.4 Mo). Mode d'accès : accès au texte intégral par intranet. Thèse de Doctorat : Électronique. Instrumentation : Alger, Ecole Nationale Polytechnique : 2026 Bibliogr. p. 97-108 Thèse confidentielle 2 ans jusqu'à Janvier 2028 Langues : Anglais (eng) Mots-clés : Behavioral biometrics Continuous authentication Multi-modal dataset Self-supervised learning Index. décimale : D000126 Résumé : Smartphone devices have become essential for managing sensitive operations such as on-line banking, accessing medical records, making digital payments, and using government services. As a result, consumers have raised the demand for robust and user-friendly authentication systems. However, traditional smartphone authentication methods such as knowledge-based (e.g., PIN codes, passwords) and static biometric systems (e.g., finger-print, facial recognition) suffer from significant limitations. These methods are vulnerable to smudge and spoofing attacks, and fails to provide ongoing security once a device is un-locked. To address these challenges, we propose the design and evaluation of efficient, multi-modal continuous authentication systems leveraging behavioral biometrics. As a starting point, the research presents a comprehensive review of state-of-the-art meth ods for continuous authentication, highlighting recent progress in behavioral biometric modalities such as hand movement and touch gestures, and identifying limitations in existing datasets. Particularly the lack of synchronized multi-modal behavioral biometric modalities, combined with static biometric characteristics such as facial features. Following, an efficient continuous authentication system is designed, by investigating advanced feature selection to identify most relevant features, showing consistent improvement with the subset of selected features. Further a new continuous authentication system is introduced using self-supervised contrastive learning. The system employs a lightweight convolutional neural network architecture based on depthwise separable convolutions, achieving high accuracy verification and identification tasks while maintaining computational efficiency. By extending the original self-supervised contrastive learning framework introduced in the previous contribution, a multi-modal fusion framework is designed, by combining hand movement and touch gesture data. Fusion is performed at the feature-level, demonstrating robust performance even on small annotated datasets where labeled biometric data is scarce. Lastly, a novel multi-modal dataset, MM-BioSync, is introduced to address the lack of synchronized behavioral biometric public datasets. The dataset integrates data from frontfacing cameras, motion sensors, and touchscreen interactions. Experiments conducted on the dataset reveal that integrating all modalities captured during reading and writing tasks yields the best performance in user verification, underscoring the value of multimodal approaches for continuous authentication. The findings of this thesis highlight the potential of behavioral biometrics and multimodal fusion in enabling continuous user authentication. By addressing key challenges related to model performance, computational efficiency, and dataset availability, this work advances the state-of-the-art and provides a foundation for developing secure and user-friendly authentication solutions on smartphones.

Réservation

Réserver ce document

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité	Spécialité	Etat_Exemplaire
T000482	D000126	Ressources électroniques	Bibliothèque centrale	Thèse de Doctorat	Disponible	Electronique	Téléchargeable

Documents numériques

Confidentiel URL

Deep neural networks optimization for embedded platforms / Anouar Laouichi (2020) 

Public ISBD Titre : Deep neural networks optimization for embedded platforms Type de document : document électronique Auteurs : Anouar Laouichi, Auteur ; Abderrahim Benaouda, Auteur ; Sid-Ahmed Berrani, Directeur de thèse ; Hamza Yous, Directeur de thèse Editeur : [S.l.] : [s.n.] Année de publication : 2020 Importance : 1 fichier PDF (6.1 M) Présentation : ill. Note générale : Mode d'accès : accès au texte intégral par intranet. Mémoire de Projet de Fin d’Études : Électronique : Alger, École Nationale Polytechnique : 2020 Bibliogr. f. 85 - 92 Langues : Anglais (eng) Mots-clés : Artificial intelligence Deep Neural Embedded Systems Inference Networks Pruning Quantization Object detection Pytorch Index. décimale : PN00820 Résumé : This project deals with the optimization of Deep Neural Networks for efficientembedded inference. Network Pruning and Quantization techniques are implemented underthe PyTorch environment and benchmarked on ResNet50. The obtained results, consisting ofcompression and speed-up rates, successfully validate the feasibility and the effectiveness of theconcept. To show their practical potential, the two schemes have been applied on RetinaNetobject detector. Additionally, this work demonstrates that inference can be performed at theedge by reducing the model’s memory footprint and the processing time, resulting in reducedlatency and energy consumption as well as improved data security. Hence, new horizons ofapplications in embedded systems are opened up Deep neural networks optimization for embedded platforms [document électronique] / Anouar Laouichi, Auteur ; Abderrahim Benaouda, Auteur ; Sid-Ahmed Berrani, Directeur de thèse ; Hamza Yous, Directeur de thèse . - [S.l.] : [s.n.], 2020 . - 1 fichier PDF (6.1 M) : ill. Mode d'accès : accès au texte intégral par intranet. Mémoire de Projet de Fin d’Études : Électronique : Alger, École Nationale Polytechnique : 2020 Bibliogr. f. 85 - 92 Langues : Anglais (eng) Mots-clés : Artificial intelligence Deep Neural Embedded Systems Inference Networks Pruning Quantization Object detection Pytorch Index. décimale : PN00820 Résumé : This project deals with the optimization of Deep Neural Networks for efficientembedded inference. Network Pruning and Quantization techniques are implemented underthe PyTorch environment and benchmarked on ResNet50. The obtained results, consisting ofcompression and speed-up rates, successfully validate the feasibility and the effectiveness of theconcept. To show their practical potential, the two schemes have been applied on RetinaNetobject detector. Additionally, this work demonstrates that inference can be performed at theedge by reducing the model’s memory footprint and the processing time, resulting in reducedlatency and energy consumption as well as improved data security. Hence, new horizons ofapplications in embedded systems are opened up

Réservation

Réserver ce document

Exemplaires (1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité	Spécialité	Etat_Exemplaire
EP00078	PN00820	Ressources électroniques	Bibliothèque centrale	Projet Fin d'Etudes	Disponible	Electronique	Téléchargeable

Documents numériques

LAOUICHI.Anouar_BENAOUDA.Abderrahim.pdf URL

Enhancing deep learning based classifiers using out of distribution data detection / Abdelghani Halimi (2022) 

Permalink

Évaluation des techniques de génération de deepfake / Achref Djaber (2024) 

Permalink

Navigation visuelle autonome de drone en milieu fermé / Mohamed Nabil Tchoulak (2020) 

Permalink

Spiking neural networks optimization / Hiba-El-Batoul Ferhat (2022) 

Permalink

The impact of the new image compression scheme JPEG AI on image analysis tasks / Mohamed Riadh Temmar (2023) 

Permalink