Advancing privacy and security in machine learning through homomorphic encryption and explainable AI
| dc.authorid | 0000-0002-8637-5473 | |
| dc.contributor.advisor | Çeliktaş, Barış | en_US |
| dc.contributor.author | Abou Harb, Mhd Raja | en_US |
| dc.contributor.other | Işık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Doktora Programı | en_US |
| dc.contributor.other | Işık University, School of Graduate Studies, Ph.D. in Computer Engineering | en_US |
| dc.date.accessioned | 2026-04-22T12:19:24Z | |
| dc.date.available | 2026-04-22T12:19:24Z | |
| dc.date.issued | 2026-03-05 | |
| dc.department | Işık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Doktora Programı | en_US |
| dc.department | Işık University, School of Graduate Studies, Ph.D. in Computer Engineering | en_US |
| dc.description | Text in English ; Abstract: English and Turkish | en_US |
| dc.description | Includes bibliographical references (leaves 111-124) | en_US |
| dc.description | xiii, 129 leaves | en_US |
| dc.description.abstract | The importance of data privacy in cloud-based Machine Learning is paramount, particularly in sectors such as healthcare and finance. Balancing robust privacy protection with high model accuracy remains a significant challenge. In this study, we propose a privacy-preserving framework utilizing ANNs on homomorphically encrypted data. To mitigate the computational complexity of non-linear activation functions (Sigmoid and Tanh), we developed lightweight, ANN-based estimators specifically designed for encrypted environments. Our experimental results demonstrate that these estimators significantly outperform traditional polynomial and piecewise linear methods, reducing MSE by up to 96% while improving accuracy and F1-scores. Our method achieved 97.70% accuracy and 0.9997 AUC on the MNIST dataset, validating its effectiveness. In real-world applications, we applied the approach to dyslexia detection using QEEG data, observing only minor performance degradation (2.66% accuracy, 3.86% AUC) compared to plaintext inference. Furthermore, a case study on the UCI Heart Disease dataset yielded 85.25% accuracy in encrypted inference, matching plaintext performance. Finally, we integrated the SHAP algorithm to ensure transparency for encrypted outputs. Our findings confirm that this approach successfully balances privacy, performance, and explainability, making it highly suitable for sensitive ML applications. | en_US |
| dc.description.abstract | Bulut tabanlı makine öğrenimi çözümlerinde, özellikle sağlık ve finans gibi hassas alanlarda veri gizliliği kritik bir öneme sahiptir. Gizlilik koruması ile yüksek model başarımı arasında denge kurmak ise güncel bir zorluktur. Bu çalışmada, homomorfik şifreleme yöntemleriyle korunan veriler üzerinde çalışan, gizlilik odaklı bir Yapay Sinir Ağı (YSA) yaklaşımı öneriyoruz. Şifreli ortamlarda hesaplama maliyeti yüksek olan Sigmoid ve Tanh gibi doğrusal olmayan aktivasyon fonksiyonlarını verimli yönetmek amacıyla, hafif YSA tabanlı tahminciler geliştirdik. Elde edilen sonuçlar, önerdiğimiz tahmincilerin geleneksel polinom ve parçalı doğrusal yöntemlere göre üstün olduğunu; Ortalama Karesel Hatayı (MSE) %96 oranında azalttığını göstermektedir. MNIST veri kümesinde elde edilen %97,70 doğruluk ve 0,9997 AUC değerleri, yöntemin etkinliğini kanıtlamıştır. Gerçek dünya senaryolarında, QEEG verileriyle disleksi tespitinde düz metin çıkarımına kıyasla ihmal edilebilir performans kayıplarıyla (%2,66 doğruluk, %3,86 AUC) başarı sağlanmıştır. UCI Kalp Hastalığı veri kümesinde ise düz metin performansıyla eşdeğer %85,25 doğruluk elde edilmiştir. Ayrıca, şeffaflığı artırmak amacıyla şifreli çıkarımlara SHAP tabanlı açıklanabilirlik entegre edilmiştir. Bulgularımız, önerilen modelin gizlilik, yüksek performans ve açıklanabilirlik gereksinimlerini başarıyla dengelediğini ve hassas sektörler için güçlü bir çözüm sunduğunu ortaya koymaktadır. | en_US |
| dc.description.tableofcontents | CLOUD-BASED MACHINE LEARNING | en_US |
| dc.description.tableofcontents | VERVIEW OF PPML | en_US |
| dc.description.tableofcontents | CHALLENGES OF NON-LINEARITY IN HOMOMORPHICALLY ENCRYPTED ML MODELS | en_US |
| dc.description.tableofcontents | RIVACY-PRESERVING AND EXPLAINABLE AI IN HEALTHCARE: INSIGHTS FROM DYSLEXIA DETECTION | en_US |
| dc.description.tableofcontents | ESEARCH GAPS IN PRIVACY-PRESERVING AND EXPLAINABLE MACHINE LEARNING | en_US |
| dc.description.tableofcontents | ANNs | en_US |
| dc.description.tableofcontents | Mathematical Foundation of ANNs | en_US |
| dc.description.tableofcontents | Learning Process of ANNs | en_US |
| dc.description.tableofcontents | Activation Functions in ANNs | en_US |
| dc.description.tableofcontents | HE | en_US |
| dc.description.tableofcontents | Mathematical Foundation of CKKS | en_US |
| dc.description.tableofcontents | XAI | en_US |
| dc.description.tableofcontents | XAI Types | en_US |
| dc.description.tableofcontents | SHAP | en_US |
| dc.description.tableofcontents | CHALLENGES WITH NON-LINEAR FUNCTIONS IN HE | en_US |
| dc.description.tableofcontents | PREVIOUS SOLUTIONS FOR NONLINEARITY IN HOMOMORPHICALLY ENCRYPTED ANN | en_US |
| dc.description.tableofcontents | Polynomial Approximation | en_US |
| dc.description.tableofcontents | Piecewise Linear Approximation | en_US |
| dc.description.tableofcontents | REGULATORY FRAMEWORKS AND DATA PRIVACY | en_US |
| dc.description.tableofcontents | GDPR | en_US |
| dc.description.tableofcontents | HIPAA | en_US |
| dc.description.tableofcontents | KVKK | en_US |
| dc.description.tableofcontents | Ensuring Regulatory Compliance with Homomorphic Encryption and Explainable AI | en_US |
| dc.description.tableofcontents | MOTIVATION FOR ANN-BASED ESTIMATORS | en_US |
| dc.description.tableofcontents | OVERVIEW OF THE PROPOSED SOLUTION | en_US |
| dc.description.tableofcontents | DESIGN OF THE MAIN ANN | en_US |
| dc.description.tableofcontents | Main ANN for MNIST Classification | en_US |
| dc.description.tableofcontents | Main ANN for Dyslexia Detection | en_US |
| dc.description.tableofcontents | DESIGNS OF THE ANN ESTIMATORS | en_US |
| dc.description.tableofcontents | HOMOMORPHICALLY ENCRYPTED INFERENCE | en_US |
| dc.description.tableofcontents | EXPLAINABILITY IN PPML | en_US |
| dc.description.tableofcontents | MNIST Dataset | en_US |
| dc.description.tableofcontents | QEEG Dataset for dyslexia detection | en_US |
| dc.description.tableofcontents | HARDWARE AND SOFTWARE ENVIRONMENT | en_US |
| dc.description.tableofcontents | HOMOMORPHIC ENCRYPTION SCHEME SETTINGS | en_US |
| dc.description.tableofcontents | CASE STUDY: PREDICTING HEART DISEASE | en_US |
| dc.description.tableofcontents | EVALUATION MEASUREMENT | en_US |
| dc.description.tableofcontents | Metrics and Procedures for Estimator Assessment | en_US |
| dc.description.tableofcontents | Evaluation Metrics for Classification Models | en_US |
| dc.description.tableofcontents | Evaluation Metrics for Explainable AI under Homomorphic Encryption | en_US |
| dc.description.tableofcontents | ERFORMANCE OF STANDALONE ACTIVATION FUNCTION ESTIMATORS | en_US |
| dc.description.tableofcontents | HOMOMORPHIC ENCRYPTION INFERENCE ON THE MNIST DATASET | en_US |
| dc.description.tableofcontents | Results Using Sigmoid Activation Function Estimators | en_US |
| dc.description.tableofcontents | Results Using Tanh Activation Function Estimators | en_US |
| dc.description.tableofcontents | Encrypted Inference Time | en_US |
| dc.description.tableofcontents | Comparative Analysis | en_US |
| dc.description.tableofcontents | REAL-WORLD APPLICATION: HE INFERENCE AND EXPLAINABLE AI FOR DYSLEXIA DETECTION | en_US |
| dc.description.tableofcontents | CASE STUDY RESULTS: HEART DISEASE PREDICTION | en_US |
| dc.description.tableofcontents | SECURITY THREAT MODEL AND GAME-THEORETIC ANALYSIS | en_US |
| dc.description.tableofcontents | Game-Theoretic Security Framework | en_US |
| dc.description.tableofcontents | Mitigation Strategies | en_US |
| dc.description.tableofcontents | ANALYSIS OF STANDALONE ACTIVATION FUNCTION ESTIMATORS | en_US |
| dc.description.tableofcontents | EVALUATING HOMOMORPHIC ENCRYPTION INFERENCE ON THE MNIST DATASET | en_US |
| dc.description.tableofcontents | NSIGHTS FROM HE INFERENCE AND EXPLAINABLE AI IN DYSLEXIA DETECTION USING QEEG DATA | en_US |
| dc.description.tableofcontents | Dyslexia Classification Discussion | en_US |
| dc.description.tableofcontents | Analysis of SHAP Perturbation Sensitivity in Encrypted Inference | en_US |
| dc.description.tableofcontents | SHAP Analysis and Neurophysiological Interpretability | en_US |
| dc.description.tableofcontents | Implications for Research and Clinical Applications | en_US |
| dc.description.tableofcontents | CASE STUDY ANALYSIS: UCI HEART DISEASE DATASET | en_US |
| dc.identifier.citation | Abou Harb, M. R. (2026). Advancing privacy and security in machine learning through homomorphic encryption and explainable AI. İstanbul: Işık Üniversitesi Lisansüstü Eğitim Enstitüsü. | en_US |
| dc.identifier.uri | https://hdl.handle.net/11729/7336 | |
| dc.institutionauthor | Abou Harb, Mhd Raja | en_US |
| dc.institutionauthorid | 0000-0002-8637-5473 | |
| dc.language.iso | en | en_US |
| dc.publisher | Işık Üniversitesi, Lisansüstü Eğitim Enstitüsü | en_US |
| dc.relation.publicationcategory | Tez | en_US |
| dc.rights | info:eu-repo/semantics/openAccess | en_US |
| dc.subject | Privacy-preserving machine learning | en_US |
| dc.subject | Homomorphic encryption | en_US |
| dc.subject | Artificial neural networks | en_US |
| dc.subject | Explainable AI | en_US |
| dc.subject | SHAP | en_US |
| dc.subject | Encrypted inference | en_US |
| dc.subject | Gizlilik koruyucu makine öğrenimi | en_US |
| dc.subject | Homomorfik şifreleme | en_US |
| dc.subject | Yapay sinir ağları | en_US |
| dc.subject | Açıklanabilir yapay zekâ | en_US |
| dc.subject | Şifreli çıkarım | en_US |
| dc.title | Advancing privacy and security in machine learning through homomorphic encryption and explainable AI | en_US |
| dc.title.alternative | Makine öğreniminde gizliliğin ve güvenliğin ilerletilmesi: homomorfik şifreleme ve açıklanabilir yapay zekâ üzerine bir çalışma | en_US |
| dc.type | Doctoral Thesis | en_US |
| dspace.entity.type | Publication | en_US |
Dosyalar
Orijinal paket
1 - 1 / 1
Yükleniyor...
- İsim:
- Advancing_privacy_and_security_in_machine_learning_through_homomorphic_encryption_and_explainable_AI.pdf
- Boyut:
- 2.57 MB
- Biçim:
- Adobe Portable Document Format
Lisans paketi
1 - 1 / 1
Küçük Resim Yok
- İsim:
- license.txt
- Boyut:
- 1.17 KB
- Biçim:
- Item-specific license agreed upon to submission
- Açıklama:












