AI-Driven Privacy Orchestration for Federated Query Processing Across Distributed Data
Silos
As organizations increasingly rely on artificial intelligence to extract insights from distributed and sensitive data, privacy-preserving data analytics has become a foundational challenge. Federated query processing across multiple private data silos—such as healthcare, finance, government, and critical infrastructure—requires mechanisms that balance data utility, privacy guarantees, system performance, and regulatory compliance. However, selecting the appropriate privacy-preserving mechanism remains difficult due to the growing design space, which includes secure multi-party computation (SMC), differential privacy (DP), hybrid SMC–DP approaches, data obfuscation, and federated learning techniques.
This presentation introduces an AI-driven privacy orchestration framework that rethinks how
privacy is specified, optimized, and enforced in federated analytics systems. Rather than requiring users to manually select low-level privacy mechanisms, the proposed approach enables declarative, intent-based privacy specification, allowing users to define what private information must be protected instead of how it should be protected. At the core of the framework is an AI-based decision engine that leverages deep learning models trained using Differentially-Private Stochastic Gradient Descent (DP-SGD). These models are used to intelligently replace selected portions of sensitive data during query execution, enabling
accurate query answering while preserving strong privacy guarantees. A cost-aware AI model
automatically reasons over accuracy loss, privacy budgets, execution latency, and storage
overhead to select optimal privacy-preserving mechanisms and tune hyperparameters dynamically. Beyond automation, the framework emphasizes human-in-the-loop AI governance. Domain experts and compliance officers can inspect, audit, and refine AI-selected privacy strategies to meet regulatory, ethical, and organizational constraints. This hybrid AI–human workflow supports transparency, explainability, and trust—key requirements for responsible and trustworthy AI deployment.
The presentation highlights how AI can move privacy protection from static rule-based systems
toward adaptive, learning-driven privacy management, enabling scalable and compliant analytics over sensitive, distributed datasets. The work is particularly relevant to conferences focused on AI systems, trustworthy AI, data management, privacy engineering, and AI governance.
Dr. Deepti Gupta is an Assistant Professor at Texas A&M University-Central Texas. After receiving her PhD, she joined Goldman Sachs as a Cloud Security Architect. She also
worked as a faculty member in the Department of Computer Science at Huston-Tillotson
University, Austin. She received her Ph.D. degree in Computer Science from the University of
Texas at San Antonio (UTSA) and also received her M.S. degree in Computer Science from UTSA.
She has worked as an Adjunct Faculty in the Department of Computer Science at St. Edward
University, Austin. Dr. Gupta’s research interests lie in the areas of security and privacy in the
Internet of Things (IoT) leveraging cloud and edge computing. Her research interests also include the application of AI and Machine Learning to secure IoT and CPS infrastructures in various application domains, such as smart healthcare, wearable IoT and smart home. She is also interested in designing federated learning algorithms to deal with non-IID data using game theory. She also developed novel anomaly detection models and fine-grained access control models to develop secure infrastructure for IoT. She has several conference and journal publications, and also continually serves as an expert reviewer for various journals and technical program committees for several conferences and workshops. Dr. Gupta has received National Science Foundation (NSF) award $349.009 "A Database Architecture for Enhanced Privacy in Machine-Learning Applications", and she also has received a Department of Defense (DoD) sub-award from Tennessee Technological University, Title : Diverse Cybersecurity- AI Capacity Building (DCAEC) Program for MSIs and Transitioning Military, Award: $124,124. She is an active team member of IEEE ComSoc Young Professionals, AnitaB.org, WiCyS, and also co-chair of the N2Women fellowship.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.