Search papers, labs, and topics across Lattice.
This paper introduces FedUP, a one-shot federated unlearning framework that addresses the challenges of knowledge loss and latency in decentralized systems by employing lightweight pluggable filters. By utilizing differentially private centroid samples to train these filters while keeping the original model parameters frozen, FedUP significantly reduces unlearning latency from minutes to seconds without compromising model performance. Experimental results across various image and text tasks show that FedUP achieves high precision in unlearning while minimizing non-target knowledge loss, demonstrating its effectiveness and efficiency in practical applications.
Unlearning in federated systems can now be achieved in seconds without sacrificing model integrity, thanks to a novel filter-based approach.
Federated unlearning (FU) is critical for complying with legal mandates like the right to be forgotten in decentralized systems, yet current methods face a persistent dilemma between non-target knowledge loss and high request latency. To resolve these issues, we propose FedUP, a one-shot federated unlearning framework utilizing lightweight pluggable filters that act as a "knowledge funnel" to screen out target data while preserving original model performance. By freezing original model parameters and training filters at the server side using differentially private (DP)-protected class centroid samples, FedUP bypasses the need for multi-round client-server communication and complex retraining, reducing unlearning latency from minutes to mere seconds. Additionally, the framework's pluggable architecture ensures inherent reversibility, enabling the seamless restoration of forgotten knowledge by simply removing the filters. Extensive experiments on diverse image and text tasks demonstrate that FedUP effectively reduces non-target knowledge loss and achieves superior unlearning precision and efficiency across various scenarios. Code is available at: https://github.com/suows/FedUP-code.
