---
title: "Proactive Federated Unlearning: Sensitivity-Guided Sparse Adaptation on Key Layers"
authors:
  - "Jinshan Lai"
  - "Fengchun Zhang"
  - "Yunyuan Wang"
  - "Dongfen Li"
  - "Yang Zhang"
  - "Xiong Li"
  - "Ruijin Wang"
doi: "10.24963/ijcai.2026/494"
url: "https://ijcai.org/proceedings/2026/494"
pdf: "https://ijcai.org/proceedings/2026/0494.pdf"
venue: "Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence"
year: 2026
pages: "4439-4447"
track: "Main Track"
keywords:
  - "Machine Learning: Federated learning"
---

# Proactive Federated Unlearning: Sensitivity-Guided Sparse Adaptation on Key Layers

Jinshan Lai, Fengchun Zhang, Yunyuan Wang, Dongfen Li, Yang Zhang, Xiong Li, Ruijin Wang

*Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence*, pages 4439-4447. DOI: [10.24963/ijcai.2026/494](https://doi.org/10.24963/ijcai.2026/494)

## Abstract

Driven by privacy regulations, federated unlearning (FU) aims to remove the influence of specific clients or samples from a trained federated model, approximating the behavior of retraining from scratch without the target data. However, existing FU methods are largely reactive: retraining-based solutions are accurate but prohibitively expensive, while parameter-based approaches are more efficient yet may cause irreversible knowledge damage and catastrophic forgetting. We introduce PFU-SKLA, a proactive FU framework that endows models with built-in forgettability. During pretraining, we perform Orthogonality-guided Representation Disentanglement (ORD) to learn a robust, disentangled feature space that reduces cross-client/class interference. Additionally, we use Dynamic Sensitivity-based Key Layer Identification (DS-KLI) to identify sensitive layers that hold target knowledge. During unlearning, we propose a sparse, lightweight adaptation strategy that precisely erases target knowledge by inserting sparse adapters into the identified critical layers while preserving non-target knowledge by freezing the backbone. Extensive experiments across multiple datasets and three standard unlearning settings show that PFU-SKLA consistently approaches retraining-from-scratch performance, while substantially reducing communication and computation costs compared to state-of-the-art (SOTA) FU methods.