Falcon: A Privacy-Preserving, Efficient, and Incentive-Aware Federated Learning Platform

Data collaboration enables multiple parties to pool data for deriving meaningful data insights. Federated learning (FL) is an emerging paradigm to facilitate data collaboration. It allows multiple parties to train a model on their joint data and make predictions without disclosing their private data to abide by privacy regulations.

Falcon is a privacy-preserving, efficient, and incentive-aware federated learning platform. It focuses on cross-silo data collaboration. We consider privacy protection the most important and design a set of privacy-preserving machine learning algorithms based on advanced techniques such as partially homomorphic encryption and secure multi-party computation. Also, since the data is distributed across different parties, we devise efficient solutions to reduce communication costs and speed up the computations. Furthermore, we propose an incentive-aware data collaboration framework to incentivize the parties to participate in the collaboration and obtain a fair reward for their contributed data.

Research Highlights

Falcon Overview

Falcon is currently designed for vertical federated learning (VFL), allowing multiple parties to train machine learning (ML) models without disclosing their raw data. First, it supports VFL training and prediction with strong and efficient privacy protection for a wide range of ML models. The protection is achieved by a hybrid strategy of threshold partially homomorphic encryption (PHE) and additive secret sharing scheme (SSS). Second, it facilitates understanding of VFL model predictions by a flexible and privacy-preserving interpretability framework, which enables the implementation of state-of-the-art interpretable methods in a decentralized setting. Third, it supports efficient data parallelism of VFL tasks and optimizes the parallelism factors to reduce the overall execution time.

Finance Example

Examplary Application A motivating example based on the our fintech could be in a car insurance application. For example, the insurance company can use AI, federated learning to conduct personalized analytics of car owners, according to their car information, driving habits, financial condition, etc., and provide a customized car insurance price for them. Meanwhile, the insurance information can be stored and maintained on a blockchain, and they can use smart contracts to automatically complete the payment and insurance claim, making the whole transaction process more efficient, secure, and reliable.


Professors and Researchers at NUS


Ooi Beng Chin

Lee Kong Chian Centennial Professor, National University of Singapore


Xiaokui Xiao

Professor, National University of Singapore


Cong Yue

Research Fellow in the Database Systems Research Group at NUS


Naili Xing

PhD Student, NUS School of Computing

International Collaborators


Yuncheng Wu

Associate Professor, Renmin University of China


Gang Chen

Professor, Zhejiang University


Meihui Zhang

Professor, Beijing Institute of Technology


Tien Tuan Anh Dinh

Senior Lecturer, Deakin University


Zhaojing Luo

Professor, Beijing Institute of Technology