Titre :	Differential Privacy in Bayesian Networks for Synthetic Data
Type de document :	document électronique
Auteurs :	Anis Louail ; Aliouat ,Zibouda, Directeur de thèse
Editeur :	Setif:UFA
Année de publication :	2025
Importance :	1 vol (56 f .)
Format :	29 cm
Langues :	Anglais (eng)
Catégories :	Thèses & Mémoires:Informatique
Mots-clés :	Informatique Differential Privacy
Index. décimale :	004 Informatique
Résumé :	With data-driven decision-making becoming increasingly prevalent, maintaining the confidentiality of protected information that resides in database systems is now an acute issue. Classical privacy-preserving methods, including Differential Privacy (DP), provide firm theoretical assurance by injecting noise that is calibrated into the answer that is output by each query. Nevertheless, such protection usually comes at the price of utility, at least where the level of noise is preset and independent of user background knowledge. At the same time, inference attacks are perilous by virtue of allowing adversaries to infer covert information by tracking a series of apparently innocuous queries. Existing inference-detecting mechanisms attempt to model and observe user knowledge but lack any effective control measures for probabilistically handling information leakage. This dissertation studies a hybrid solution that blends Differential Privacy with user-adapted knowledge modeling for adaptively regulating the level of privacy. With probabilistic graphical models, notably Bayesian networks, we want to predict what the user would infer from the response to queries. The level of noise due to DP is then fixed based on this estimated knowledge such that utility and privacy are traded off. Our approach is validated by an implementation prototype with a PostgreSQL database, populated with inference monitoring and DP protection. The result demonstrates utility growth with stronger privacy protection by varying protection with user knowledge.
Note de contenu :	Sommaire 1 Introduction and Problem Statement 6 1.1 General Introduction . . . . . . . . . . . . . . . . . . . . . . . 6 1.2 Real-World Motivation . . . . . . . . . . . . . . . . . . . . . . 7 1.3 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . 7 1.4 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . 8 1.5 Methodological Approach . . . . . . . . . . . . . . . . . . . . 8 1.6 Structure of the Thesis . . . . . . . . . . . . . . . . . . . . . . 9 2 State of the Art 10 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.2 Data Privacy in Databases . . . . . . . . . . . . . . . . . . . . 11 2.2.1 Overview of Privacy Concerns . . . . . . . . . . . . . . 11 2.2.2 Traditional Privacy Techniques . . . . . . . . . . . . . 11 2.2.3 Limitations of Conventional Approaches . . . . . . . . 12 2.3 Inference Attacks . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.3.1 Definition and Motivation . . . . . . . . . . . . . . . . 12 2.3.2 Types of Inference Attacks . . . . . . . . . . . . . . . . 13 2.3.3 Inference Attacks in Non-Interactive Settings . . . . . . 13 2.4 Bayesian Networks in Privacy and Security . . . . . . . . . . . 14 2.4.1 Bayesian Networks Overview . . . . . . . . . . . . . . . 14 2.4.2 Why Use Bayesian Networks in Privacy Research . . . 15 2.4.3 Use in Modeling Inference . . . . . . . . . . . . . . . . 16 2.4.4 Applications . . . . . . . . . . . . . . . . . . . . . . . . 17 2.5 Differential Privacy Mechanisms . . . . . . . . . . . . . . . . . 18 2.5.1 Motivation and Formal Guarantee . . . . . . . . . . . . 18 2.5.2 Global vs. Local Differential Privacy . . . . . . . . . . 19 2.5.3 Mechanisms for Ensuring Differential Privacy . . . . . 19 2.5.4 Comparison of DP Mechanisms . . . . . . . . . . . . . 20 2.6 Evaluation Metrics for Differentially Private Synthetic Data . 20 2.6.1 On the Notion of Fidelity . . . . . . . . . . . . . . . . 21 2.7 Selected Methods for Differentially Private Tabular Data Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.7.1 PrivBayes . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.7.2 MWEM (Multiplicative Weights Exponential Mechanism) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.7.3 DualQuery . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.7.4 Private-PGM . . . . . . . . . . . . . . . . . . . . . . . 23 2.7.5 Junction Tree Method . . . . . . . . . . . . . . . . . . 23 2.7.6 Justification of Selected Comparisons . . . . . . . . . . 24 2.7.7 Overview of PrivBayes . . . . . . . . . . . . . . . . . . 24 2.7.8 MWEM (Multiplicative Weights Exponential Mechanism) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.7.9 DualQuery . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.7.10 Private-PGM . . . . . . . . . . . . . . . . . . . . . . . 26 2.7.11 Junction Tree Methods . . . . . . . . . . . . . . . . . . 26 2.8 In-depth Comparison of PrivBayes and Private-PGM . . . . . 27 2.8.1 Theoretical Foundations . . . . . . . . . . . . . . . . . 27 2.8.2 Computational Complexity and Efficiency . . . . . . . 27 2.8.3 Handling of Different Data Types . . . . . . . . . . . . 27 2.8.4 Quality and Accuracy of Generated Data . . . . . . . . 28 2.8.5 Practical Applications and Recommendations . . . . . 28 2.9 In-depth Comparison of PrivBayes and Junction Tree Methods 28 2.9.1 Fundamental Differences . . . . . . . . . . . . . . . . . 28 2.9.2 Computational Efficiency and Complexity . . . . . . . 29 2.9.3 Inference Mechanisms . . . . . . . . . . . . . . . . . . 29 2.9.4 Handling of Data Types and Structure . . . . . . . . . 29 2.9.5 Privacy Budget Utilization . . . . . . . . . . . . . . . . 29 2.9.6 Synthetic Data Accuracy and Generalizability . . . . . 30 2.9.7 Practical Recommendations . . . . . . . . . . . . . . . 30 2.10 Related Work on DP Database Publishing . . . . . . . . . . . 30 2.11 Limitations and Research Gaps . . . . . . . . . . . . . . . . . 35 2.11.1 Summary of Limitations . . . . . . . . . . . . . . . . . 35 2.11.2 Motivation for Our Work . . . . . . . . . . . . . . . . . 36 2.12 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3 Methodology 38 3.1 Overview of the Proposed Approach . . . . . . . . . . . . . . . 38 3.2 Bayesian Network Structure Learning . . . . . . . . . . . . . . 38 3.3 Integration of Differential Privacy . . . . . . . . . . . . . . . . 39 3.4 Synthetic Data Generation . . . . . . . . . . . . . . . . . . . . 40 3.5 Evaluation Strategy . . . . . . . . . . . . . . . . . . . . . . . . 40 3.5.1 Structure Comparison . . . . . . . . . . . . . . . . . . 41 3.5.2 Data Utility . . . . . . . . . . . . . . . . . . . . . . . . 41 3.6 Structural Repair of DP Networks . . . . . . . . . . . . . . . . 41 3.6.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.6.2 Edge Injection Method . . . . . . . . . . . . . . . . . . 42 3.6.3 Synthetic Data After Structural Repair . . . . . . . . . 42 4 Results and Discussion 43 4.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . 43 4.1.1 Dataset Description . . . . . . . . . . . . . . . . . . . . 43 4.1.2 Tools and Libraries . . . . . . . . . . . . . . . . . . . . 44 4.1.3 Parameter Settings . . . . . . . . . . . . . . . . . . . . 44 4.2 Utility and Structural Similarity Evaluation . . . . . . . . . . 45 4.2.1 KL Divergence Across Variables . . . . . . . . . . . . . 45 4.2.2 Bayesian Network Structure Evaluation . . . . . . . . . 46 4.2.3 Classification Utility . . . . . . . . . . . . . . . . . . . 48 4.3 Structural Repair and Its Impact . . . . . . . . . . . . . . . . 48 4.3.1 Edge Injection Experiments . . . . . . . . . . . . . . . 48 4.3.2 Utility After Repair . . . . . . . . . . . . . . . . . . . . 50 4.4 Comparison Between Methods . . . . . . . . . . . . . . . . . . 51 4.5 Discussion and Limitations . . . . . . . . . . . . . . . . . . . . 51 General Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Côte titre :	MAI/1039

Differential Privacy in Bayesian Networks for Synthetic Data [document électronique] / Anis Louail ; Aliouat ,Zibouda, Directeur de thèse . - [S.l.] : Setif:UFA, 2025 . - 1 vol (56 f .) ; 29 cm.
Langues : Anglais (eng)

Catégories :	Thèses & Mémoires:Informatique
Mots-clés :	Informatique Differential Privacy
Index. décimale :	004 Informatique
Résumé :	With data-driven decision-making becoming increasingly prevalent, maintaining the confidentiality of protected information that resides in database systems is now an acute issue. Classical privacy-preserving methods, including Differential Privacy (DP), provide firm theoretical assurance by injecting noise that is calibrated into the answer that is output by each query. Nevertheless, such protection usually comes at the price of utility, at least where the level of noise is preset and independent of user background knowledge. At the same time, inference attacks are perilous by virtue of allowing adversaries to infer covert information by tracking a series of apparently innocuous queries. Existing inference-detecting mechanisms attempt to model and observe user knowledge but lack any effective control measures for probabilistically handling information leakage. This dissertation studies a hybrid solution that blends Differential Privacy with user-adapted knowledge modeling for adaptively regulating the level of privacy. With probabilistic graphical models, notably Bayesian networks, we want to predict what the user would infer from the response to queries. The level of noise due to DP is then fixed based on this estimated knowledge such that utility and privacy are traded off. Our approach is validated by an implementation prototype with a PostgreSQL database, populated with inference monitoring and DP protection. The result demonstrates utility growth with stronger privacy protection by varying protection with user knowledge.
Note de contenu :	Sommaire 1 Introduction and Problem Statement 6 1.1 General Introduction . . . . . . . . . . . . . . . . . . . . . . . 6 1.2 Real-World Motivation . . . . . . . . . . . . . . . . . . . . . . 7 1.3 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . 7 1.4 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . 8 1.5 Methodological Approach . . . . . . . . . . . . . . . . . . . . 8 1.6 Structure of the Thesis . . . . . . . . . . . . . . . . . . . . . . 9 2 State of the Art 10 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.2 Data Privacy in Databases . . . . . . . . . . . . . . . . . . . . 11 2.2.1 Overview of Privacy Concerns . . . . . . . . . . . . . . 11 2.2.2 Traditional Privacy Techniques . . . . . . . . . . . . . 11 2.2.3 Limitations of Conventional Approaches . . . . . . . . 12 2.3 Inference Attacks . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.3.1 Definition and Motivation . . . . . . . . . . . . . . . . 12 2.3.2 Types of Inference Attacks . . . . . . . . . . . . . . . . 13 2.3.3 Inference Attacks in Non-Interactive Settings . . . . . . 13 2.4 Bayesian Networks in Privacy and Security . . . . . . . . . . . 14 2.4.1 Bayesian Networks Overview . . . . . . . . . . . . . . . 14 2.4.2 Why Use Bayesian Networks in Privacy Research . . . 15 2.4.3 Use in Modeling Inference . . . . . . . . . . . . . . . . 16 2.4.4 Applications . . . . . . . . . . . . . . . . . . . . . . . . 17 2.5 Differential Privacy Mechanisms . . . . . . . . . . . . . . . . . 18 2.5.1 Motivation and Formal Guarantee . . . . . . . . . . . . 18 2.5.2 Global vs. Local Differential Privacy . . . . . . . . . . 19 2.5.3 Mechanisms for Ensuring Differential Privacy . . . . . 19 2.5.4 Comparison of DP Mechanisms . . . . . . . . . . . . . 20 2.6 Evaluation Metrics for Differentially Private Synthetic Data . 20 2.6.1 On the Notion of Fidelity . . . . . . . . . . . . . . . . 21 2.7 Selected Methods for Differentially Private Tabular Data Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.7.1 PrivBayes . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.7.2 MWEM (Multiplicative Weights Exponential Mechanism) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.7.3 DualQuery . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.7.4 Private-PGM . . . . . . . . . . . . . . . . . . . . . . . 23 2.7.5 Junction Tree Method . . . . . . . . . . . . . . . . . . 23 2.7.6 Justification of Selected Comparisons . . . . . . . . . . 24 2.7.7 Overview of PrivBayes . . . . . . . . . . . . . . . . . . 24 2.7.8 MWEM (Multiplicative Weights Exponential Mechanism) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.7.9 DualQuery . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.7.10 Private-PGM . . . . . . . . . . . . . . . . . . . . . . . 26 2.7.11 Junction Tree Methods . . . . . . . . . . . . . . . . . . 26 2.8 In-depth Comparison of PrivBayes and Private-PGM . . . . . 27 2.8.1 Theoretical Foundations . . . . . . . . . . . . . . . . . 27 2.8.2 Computational Complexity and Efficiency . . . . . . . 27 2.8.3 Handling of Different Data Types . . . . . . . . . . . . 27 2.8.4 Quality and Accuracy of Generated Data . . . . . . . . 28 2.8.5 Practical Applications and Recommendations . . . . . 28 2.9 In-depth Comparison of PrivBayes and Junction Tree Methods 28 2.9.1 Fundamental Differences . . . . . . . . . . . . . . . . . 28 2.9.2 Computational Efficiency and Complexity . . . . . . . 29 2.9.3 Inference Mechanisms . . . . . . . . . . . . . . . . . . 29 2.9.4 Handling of Data Types and Structure . . . . . . . . . 29 2.9.5 Privacy Budget Utilization . . . . . . . . . . . . . . . . 29 2.9.6 Synthetic Data Accuracy and Generalizability . . . . . 30 2.9.7 Practical Recommendations . . . . . . . . . . . . . . . 30 2.10 Related Work on DP Database Publishing . . . . . . . . . . . 30 2.11 Limitations and Research Gaps . . . . . . . . . . . . . . . . . 35 2.11.1 Summary of Limitations . . . . . . . . . . . . . . . . . 35 2.11.2 Motivation for Our Work . . . . . . . . . . . . . . . . . 36 2.12 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3 Methodology 38 3.1 Overview of the Proposed Approach . . . . . . . . . . . . . . . 38 3.2 Bayesian Network Structure Learning . . . . . . . . . . . . . . 38 3.3 Integration of Differential Privacy . . . . . . . . . . . . . . . . 39 3.4 Synthetic Data Generation . . . . . . . . . . . . . . . . . . . . 40 3.5 Evaluation Strategy . . . . . . . . . . . . . . . . . . . . . . . . 40 3.5.1 Structure Comparison . . . . . . . . . . . . . . . . . . 41 3.5.2 Data Utility . . . . . . . . . . . . . . . . . . . . . . . . 41 3.6 Structural Repair of DP Networks . . . . . . . . . . . . . . . . 41 3.6.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . 41 3.6.2 Edge Injection Method . . . . . . . . . . . . . . . . . . 42 3.6.3 Synthetic Data After Structural Repair . . . . . . . . . 42 4 Results and Discussion 43 4.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . 43 4.1.1 Dataset Description . . . . . . . . . . . . . . . . . . . . 43 4.1.2 Tools and Libraries . . . . . . . . . . . . . . . . . . . . 44 4.1.3 Parameter Settings . . . . . . . . . . . . . . . . . . . . 44 4.2 Utility and Structural Similarity Evaluation . . . . . . . . . . 45 4.2.1 KL Divergence Across Variables . . . . . . . . . . . . . 45 4.2.2 Bayesian Network Structure Evaluation . . . . . . . . . 46 4.2.3 Classification Utility . . . . . . . . . . . . . . . . . . . 48 4.3 Structural Repair and Its Impact . . . . . . . . . . . . . . . . 48 4.3.1 Edge Injection Experiments . . . . . . . . . . . . . . . 48 4.3.2 Utility After Repair . . . . . . . . . . . . . . . . . . . . 50 4.4 Comparison Between Methods . . . . . . . . . . . . . . . . . . 51 4.5 Discussion and Limitations . . . . . . . . . . . . . . . . . . . . 51 General Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Côte titre :	MAI/1039