pFedUL: Layer-Aware Federated Unlearning for Personalized Federated Learning
2026-06-15 • Machine Learning
Machine Learning
AI summaryⓘ
The authors study how to remove a specific user's data from personalized federated learning models, which combine common global parts with individual personalized parts. They point out a challenge: fully erasing a user's data could harm the unique model parts for other users. To solve this, they create a method called pFedUL that carefully assesses how much each layer was influenced by the user and selectively forgets that data while keeping personalization intact. Their method also helps other users quickly adjust their models afterward. In tests on image datasets, their approach effectively forgets data like full retraining but better preserves personalization compared to existing methods.
federated learningfederated unlearningpersonalized federated learningnon-IID datamodel layersFedAvgFedPercontribution attributionpersonalization preservationGDPR
Authors
Zhuodong Liu, Xiangyu Li, Zhihao Zhang
Abstract
Federated unlearning (FU) enables the removal of specific data contributions from federated learning (FL) models to comply with regulations such as the General Data Protection Regulation (GDPR). However, most existing FU methods are designed for the FedAvg paradigm, where all clients share a single global model. In practice, personalized federated learning (pFL) methods such as FedPer, FedRep, Ditto, and FedBN have become widely adopted due to their superior handling of non-IID data. These methods decompose the model into shared global layers and client-specific personalized layers, fundamentally altering the semantics of unlearning, yet this setting has received little attention. We formalize FU under the pFL paradigm, identifying a tension between unlearning completeness on shared layers and personalization preservation for remaining clients. We then propose pFedUL, a layer-aware selective unlearning framework comprising three components: (1) gradient-based layer-wise contribution attribution that separately quantifies the target client's influence on shared and personalized parameters, (2) adaptive selective unlearning that applies differentiated forgetting strategies across layer types, and (3) a lightweight recalibration protocol enabling remaining clients to restore personalization with minimal overhead. We further introduce two new metrics, Personalization Preservation Score (PPS) and Cross-client Fairness Index (CFI), to evaluate pFL-specific unlearning quality. Experiments on CIFAR-10, CIFAR-100, and FEMNIST under varying non-IID settings indicate that pFedUL achieves unlearning effectiveness comparable to full retraining while maintaining an average of 97.3\% personalized accuracy for remaining clients. Compared with six state-of-the-art FU methods adapted to the pFL setting, pFedUL consistently achieves superior personalization preservation.