Enhancing Federated Learning Privacy with QUBO

📅 2025-11-04
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
In federated learning (FL), frequent client participation in model updates leads to cumulative leakage of sensitive information, rendering systems vulnerable to membership/attribute inference and model inversion attacks. To address this, we propose a quantum-inspired QUBO (Quadratic Unconstrained Binary Optimization)-based dynamic client selection mechanism—the first to formulate FL client selection as a binary optimization problem. Our method employs a validation-set-driven, fine-grained update screening strategy to suppress participation frequency of high-risk clients, thereby achieving privacy-utility co-optimization without compromising global model accuracy. Experiments on MNIST (300 clients) and CINIC-10 (30 clients) demonstrate that our approach reduces per-round privacy exposure by 95.2% (49% cumulatively) while maintaining or even exceeding the accuracy of full aggregation—despite rejecting 147 updates. On CINIC-10, per-round and cumulative privacy gains reach 82% and 33%, respectively.

Technology Category

Application Category

📝 Abstract
Federated learning (FL) is a widely used method for training machine learning (ML) models in a scalable way while preserving privacy (i.e., without centralizing raw data). Prior research shows that the risk of exposing sensitive data increases cumulatively as the number of iterations where a client's updates are included in the aggregated model increase. Attackers can launch membership inference attacks (MIA; deciding whether a sample or client participated), property inference attacks (PIA; inferring attributes of a client's data), and model inversion attacks (MI; reconstructing inputs), thereby inferring client-specific attributes and, in some cases, reconstructing inputs. In this paper, we mitigate risk by substantially reducing per client exposure using a quantum computing-inspired quadratic unconstrained binary optimization (QUBO) formulation that selects a small subset of client updates most relevant for each training round. In this work, we focus on two threat vectors: (i) information leakage by clients during training and (ii) adversaries who can query or obtain the global model. We assume a trusted central server and do not model server compromise. This method also assumes that the server has access to a validation/test set with global data distribution. Experiments on the MNIST dataset with 300 clients in 20 rounds showed a 95.2% per-round and 49% cumulative privacy exposure reduction, with 147 clients'updates never being used during training while maintaining in general the full-aggregation accuracy or even better. The method proved to be efficient at lower scale and more complex model as well. A CINIC-10 dataset-based experiment with 30 clients resulted in 82% per-round privacy improvement and 33% cumulative privacy.
Problem

Research questions and friction points this paper is trying to address.

Mitigating cumulative privacy risks in federated learning from multiple attack types
Reducing client data exposure during model training iterations using QUBO optimization
Selecting minimal client updates while maintaining model accuracy against adversaries
Innovation

Methods, ideas, or system contributions that make the work stand out.

Uses QUBO formulation for client selection
Reduces privacy exposure per training round
Maintains model accuracy with fewer client updates
🔎 Similar Papers
No similar papers found.
A
Andras Ferenczi
American Express Co
S
Sutapa Samanta
American Express Co
D
Dagen Wang
American Express Co
T
Todd Hodges
American Express Co