TADP-RME: A Trust-Adaptive Differential Privacy Framework for Enhancing Reliability of Data-Driven Systems
2026-04-09 • Cryptography and Security
Cryptography and SecurityArtificial IntelligenceMachine Learning
AI summaryⓘ
The authors address the problem of keeping data private while using it for learning, especially when users trust the system differently. They introduce TADP-RME, a method that adjusts how much privacy is applied depending on an inverse trust score, allowing a better balance between privacy and usefulness. They also use a technique called Reverse Manifold Embedding to hide geometric patterns in the data that attackers might exploit, without breaking privacy guarantees. Their approach reduces the chance of privacy attacks and keeps data useful better than previous methods.
differential privacyprivacy budgettrust scoremanifold embeddinginference attacksutility-privacy trade-offpost-processingadversarial settingsdata-driven systemsprivacy leakage
Authors
Labani Halder, Payel Sadhukhan, Sarbani Palit
Abstract
Ensuring reliability in adversarial settings necessitates treating privacy as a foundational component of data-driven systems. While differential privacy and cryptographic protocols offer strong guarantees, existing schemes rely on a fixed privacy budget, leading to a rigid utility-privacy trade-off that fails under heterogeneous user trust. Moreover, noise-only differential privacy preserves geometric structure, which inference attacks exploit, causing privacy leakage. We propose TADP-RME (Trust-Adaptive Differential Privacy with Reverse Manifold Embedding), a framework that enhances reliability under varying levels of user trust. It introduces an inverse trust score in the range [0,1] to adaptively modulate the privacy budget, enabling smooth transitions between utility and privacy. Additionally, Reverse Manifold Embedding applies a nonlinear transformation to disrupt local geometric relationships while preserving formal differential privacy guarantees through post-processing. Theoretical and empirical results demonstrate improved privacy-utility trade-offs, reducing attack success rates by up to 3.1 percent without significant utility degradation. The framework consistently outperforms existing methods against inference attacks, providing a unified approach for reliable learning in adversarial environments.