Home / Protecting privacy with deep-learning tools


Privacy-Preserving Analytics for Non-IID Data

In the era of big data, the unprecedented speed and volume of data collection poses benefits as well as risks. The large amount of collected data contains critical information for daily homeland security operations as well as potential usage to enhance analytical capabilities for better decision-making. The potential risk of using operational data for analysis across the HSE is release of sensitive and personal information. Despite the extensive research efforts dedicated to privacy-enhancing technologies over several decades, a major gap remains between those research efforts and their operational application across the enterprise.


This project is designed to create a suite of novel deep learning techniques enabling analysis of distributed non-IID datasets which are datasets held by multiple parties that do not follow the same underlying distribution. Because these types of datasets are more challenging to model there has been limited research on effectively implementing privacy-preserving techniques. This research project is approaching this challenge through two strategies, the first method uses generated data, and the second enables privacy-preserving computation without data sharing. Both strategies allow DHS components more access to using data more efficiently while protecting against unauthorized release of private information.


The outcomes from this project will provide software tools that will significantly enhance the capabilities of DHS components to derive critical insights from the large amount of data currently held by multiple parties while protecting the release of sensitive information. The impact of this project is to develop tools that will prevent the release of sensitive data to unauthorized parties and protect against data breaches.

Research Leadership Team

Principal Investigator: Jingrui He, University of Illinois at Urbana-Champaign
Co-PI: Ross Maciejewski, Arizona State University
Co-PI: Hanghang Tong, University of Illinois at Urbana-Champaign

Data analytics


Discover more projects

The CAOE is committed to developing innovative tools and techniques to safeguard our homeland from potential threats and vulnerabilities.