| Titre : | Tolérance aux pannes automatique des services composites dans l’Internet des Objets |
| Auteurs : | Siham Sahli, Auteur |
| Type de document : | Thése doctorat |
| Editeur : | BENHARZALLAH, 2026 |
| Format : | 1 vol. (281 p.) / couv. ill. en coul / 30 |
| Langues: | Anglais |
| Langues originales: | Anglais |
| Résumé : |
The Internet of Things (IoT), as a dynamic environment, increasingly integrates alarge number of devices capable of interacting with and modifying the physical world.These devices are abstracted as IoT services, which can be either simple or composite.Composite services are formed through the integration of multiple data flows generated by heterogeneous simple and/or composite services. However, these composite services are highly vulnerable to failures such as sensor outages, communication disruptions, ordata corruption. Such failures often lead to missing or inconsistent data streams that may compromise service continuity and reliability.This thesis proposes an automated fault-tolerant framework for IoT composite servicesaimed at maintaining service continuity through intelligent reconstruction of incomplete data. At the core of the framework lies a Bidirectional Long Short-Term Memory (BiLSTM) Autoencoder that captures temporal patterns in multivariate time-series data. Inaddition, a Fuzzy Logic Controller (FLC) applies domain-specific rules to makedecisions.As a second contribution, the BI-LSTM AE is enhanced through context integration to highlight relevant contextualfeatures to guide the reconstruction process under varyingenvironmental conditions. The proposed framework is evaluated on a real-world healthcare monitoring dataset.Experimental results show promising reconstruction accuracy and improved robustnesscompared with the implemented baseline configurations. The patient monitoring scenarioserves as a representative case study to illustrate the practical applicability of the proposed automated fault-tolerance framework. |
| Sommaire : |
Acknowledgement I Dedication II Abstract III Résumé IV Vيٴٰؒڞ List of Figures X List of Tables XIII List of Abbreviations XIV List of Publications XVI 1 General Introduction 3 1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 3 Main Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 4 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 VICONTENTS 2 Preliminaries and Fundamental Concepts 10 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.2 IoT Composite Services: Overview . . . . . . . . . . . . . . . . . . . . . 11 2.2.1 Internet of Things (IoT) definition . . . . . . . . . . . . . . . . . 11 2.2.2 From web services to Internet of Things services . . . . . . . . . . 12 2.2.3 Composition of IoT services . . . . . . . . . . . . . . . . . . . . . 14 2.3 Fault Tolerance of IoT Composite Services . . . . . . . . . . . . . . . . . 16 2.3.1 Faults, Errors, and Failures . . . . . . . . . . . . . . . . . . . . . 16 2.3.2 Fault Tolerance and Self-Healing Mechanisms . . . . . . . . . . . 16 2.3.3 Fault Recovery Mechanisms . . . . . . . . . . . . . . . . . . . . . 18 2.3.4 Dependability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.4 Machine Learning Overview . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.4.1 Supervised Learning . . . . . . . . . . . . . . . . . . . . . . . . . 22 2.4.2 Unsupervised Learning . . . . . . . . . . . . . . . . . . . . . . . . 24 2.4.3 Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . . . 25 2.5 Deep Learning Models for Temporal Data . . . . . . . . . . . . . . . . . 25 2.5.1 Recurrent Neural Networks (RNN) . . . . . . . . . . . . . . . . . 26 2.5.2 Long Short-Term Memory (LSTM) . . . . . . . . . . . . . . . . . 27 2.5.3 Bidirectional LSTM (Bi-LSTM) . . . . . . . . . . . . . . . . . . . 28 2.5.4 Autoencoder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2.6 Fuzzy Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2.6.1 Fuzzy Set Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 2.6.2 Membership Functions . . . . . . . . . . . . . . . . . . . . . . . . 32 2.6.3 Fuzzy Set Operations . . . . . . . . . . . . . . . . . . . . . . . . 33 2.6.4 Types of Fuzzy Inference Systems . . . . . . . . . . . . . . . . . . 35 2.6.5 Applications of Fuzzy Logic in Intelligent Systems . . . . . . . . . 35 2.6.6 Advantages of Fuzzy Logic . . . . . . . . . . . . . . . . . . . . . . 36 2.7 Data Reconstruction and Imputation Techniques . . . . . . . . . . . . . 36 2.7.1 The Problem of Missing Data in Time Series . . . . . . . . . . . . 37 VIICONTENTS 2.7.2 Missing Data Mechanisms and Patterns . . . . . . . . . . . . . . 37 2.7.3 Classical Imputation Techniques . . . . . . . . . . . . . . . . . . 38 2.7.4 Machine Learning Approaches for Data Imputation . . . . . . . . 39 2.7.5 Deep Learning-based Reconstruction Methods . . . . . . . . . . . 39 2.7.6 Autoencoder-based Time Series Reconstruction . . . . . . . . . . 40 2.7.7 LSTM Encoder–Decoder Models . . . . . . . . . . . . . . . . . . 41 2.7.8 Bidirectional LSTM Models for Data Imputation . . . . . . . . . 41 2.7.9 Advanced Models: Attention and Transformer Architectures . . . 42 2.8 Context Awareness: Overview . . . . . . . . . . . . . . . . . . . . . . . . 42 2.8.1 Introduction to Context-Aware Computing . . . . . . . . . . . . . 42 2.8.2 Definition of Context . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.8.3 Conceptual Dimensions of Context . . . . . . . . . . . . . . . . . 44 2.8.4 Context Space and Multidimensional Context Models . . . . . . . 44 2.8.5 Context Lifecycle in Context-Aware Systems . . . . . . . . . . . . 45 2.8.6 Context Representation and Context Modeling . . . . . . . . . . 46 2.8.7 Context in Smart Systems and IoT Environments . . . . . . . . . 46 2.8.8 Challenges in Context Modeling . . . . . . . . . . . . . . . . . . . 47 2.8.9 Role of Context in Data Analysis . . . . . . . . . . . . . . . . . . 47 2.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 3 State-of-the-Art Analysis 49 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.2 Overview of fault tolerance approaches . . . . . . . . . . . . . . . . . . . 50 3.2.1 Technical Paradigm . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.2.2 Implementation level . . . . . . . . . . . . . . . . . . . . . . . . . 57 3.3 Comparison of AFT approaches . . . . . . . . . . . . . . . . . . . . . . . 59 3.3.1 Reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 3.3.2 Availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 VIIICONTENTS 3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 4 Automated Fault Tolerance of IoT CS framework 67 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.2 Related work relevant to our approach . . . . . . . . . . . . . . . . . . . 68 4.3 Proposed framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4.3.1 Fault tolerant phase . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.3.2 Reconstruction loss . . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.3.3 Reconstruction error . . . . . . . . . . . . . . . . . . . . . . . . . 78 4.3.4 Threshold computation . . . . . . . . . . . . . . . . . . . . . . . 79 4.3.5 Decision phase . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 4.3.6 Detailed algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . 81 4.4 Context-aware extended BI-LSTM Autoencoder . . . . . . . . . . . . . 82 4.4.1 Context-Conditioned Gaussian Mixture . . . . . . . . . . . . . . 85 4.4.2 Contextual Gating Network . . . . . . . . . . . . . . . . . . . . . 86 4.4.3 Likelihood-Based Anomaly Score . . . . . . . . . . . . . . . . . . 87 4.4.4 Weak Supervision for Regime Learning . . . . . . . . . . . . . . . 88 4.4.5 Training Objective . . . . . . . . . . . . . . . . . . . . . . . . . . 88 4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 5 Evaluation and Experimental Results 91 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 5.2 Implementation and Experimental Setup . . . . . . . . . . . . . . . . . . 92 5.2.1 Dataset description . . . . . . . . . . . . . . . . . . . . . . . . . . 93 5.2.2 Model developement . . . . . . . . . . . . . . . . . . . . . . . . . 93 5.2.3 Dataset Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 5.2.4 Training Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 96 5.2.5 Score Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 5.2.6 Experimental Variants . . . . . . . . . . . . . . . . . . . . . . . . 97 5.2.7 Reproducibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 IXCONTENTS 5.2.8 Model Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . 97 5.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 5.3.1 Baseline BI-LSTM Autoencoder . . . . . . . . . . . . . . . . . . . 98 5.3.2 Fuzzy Logic Controller: Implementation and Evaluation . . . . . 100 5.3.3 Contextual BI-LSTM Autoencoder . . . . . . . . . . . . . . . . . 102 5.4 Discussion of Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 Conclusion and Future Works 111 1 Summary and key findings . . . . . . . . . . . . . . . . . . . . . . . . . . 112 2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 |
| Type de document : | Thése doctorat |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| TINF/217 | Théses de doctorat | bibliothèque sciences exactes | Consultable |
