Skip to main content
TRUST

Research Thrust

Secure & Privacy-Preserving Distributed Intelligence

Building machine learning and data-sharing systems that remain accurate and useful even when data, participants, or infrastructure cannot be fully trusted.

Modern intelligent systems rarely learn from a single, centralized dataset. Instead, they train and operate across distributed organizations, edge devices, sensor networks, and federated infrastructures where data ownership, privacy regulations, and adversarial participants all constrain what can be shared and how.

This thrust investigates methods for secure and privacy-preserving distributed intelligence, including:

  • Federated and distributed learning under non-IID data, communication constraints, and Byzantine or adversarial participants
  • Privacy-preserving computation techniques, including differential privacy, secure aggregation, and privacy-utility trade-off analysis
  • Trust and integrity mechanisms for multi-party and cross-organizational data pipelines
  • Resilient distributed architectures for scientific and mission-critical infrastructure

The goal is to enable organizations, scientific collaborations, and networked systems to collaboratively build and deploy intelligent models without compromising the confidentiality, integrity, or availability of the underlying data.

Related Work

Publications in this Area

Conference Submitted

Manuscript submitted — under peer review; not yet published.

FedTrace: Forensic Client Attribution and Attack-Onset Localization in Heterogeneous Federated Learning

Al Amin

NDSS 2027, 2027

AI Security, Cyber Defense & ForensicsSecure & Privacy-Preserving Distributed Intelligence
Cite this work

Al Amin. "FedTrace: Forensic Client Attribution and Attack-Onset Localization in Heterogeneous Federated Learning." Submitted to NDSS 2027.

← Back to all research areas