$740K Grant Will Develop Trustworthy Federated Learning for Generative AI

Associate Professor Hai Phan in Ying Wu College of Computing’s Department of Data Science has been awarded a $740,000 grant through the Qatar Research, Development and Innovation Council (QRDI) to develop a trustworthy federated learning (FL) model that will address new standards for AI safety and security as outlined in the White House’s Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence.

The highly competitive award is one of 18 funded out of 500+ submissions and will support ProFL, the first Private and Robust FL framework for jointly training machine learning (ML) models. The grant will fund phase one of the project, which is being done in partnership with NJIT and Qatar Computing Research Institute (QCRI).

As machine learning continues to drive technology forward, high stakeholder usage has increased by approximately 1,000% over the last decade, according to Phan. The transition to cloud-based services has necessitated recent data privacy and security regulations, which pose significant challenges in collecting and using personally sensitive data for ML applications. Federated learning is a promising solution for these challenges by allowing clients to train better ML models jointly through a coordinating server that allows sensitive data to remain with clients during federated training.

Federated learning can enable myriad applications in practice, from healthcare and crime detection to human sensing based on mobile signals, among many others, without the prohibitive cost of privacy, security and administrative procedures of centralized data integration.

“Technology has advanced to a non-coding model that enables the general public to use AI in their every day lives. However, if concerns over security and privacy prevent this, such advances are meaningless,” said Phan.

Future phases of the project, which are part of a larger grant, will support building a framework for secure generative AI that Phan has named the Einstein Trust Framework. He expects to have preliminary results on his findings next year.

Phan added, “Safe technology equals safe, confident minds… But the race is on to ensure it is useful for society.”