{"type":"ScholarlyArticle","title":"Proactive Federated Unlearning: Sensitivity-Guided Sparse Adaptation on Key Layers","authors":[{"name":"Jinshan Lai","slug":"jinshan-lai","given":"Jinshan","family":"Lai","affiliation":""},{"name":"Fengchun Zhang","slug":"fengchun-zhang","given":"Fengchun","family":"Zhang","affiliation":""},{"name":"Yunyuan Wang","slug":"yunyuan-wang","given":"Yunyuan","family":"Wang","affiliation":""},{"name":"Dongfen Li","slug":"dongfen-li","given":"Dongfen","family":"Li","affiliation":""},{"name":"Yang Zhang","slug":"yang-zhang","given":"Yang","family":"Zhang","affiliation":""},{"name":"Xiong Li","slug":"xiong-li","given":"Xiong","family":"Li","affiliation":""},{"name":"Ruijin Wang","slug":"ruijin-wang","given":"Ruijin","family":"Wang","affiliation":""}],"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.","doi":"10.24963/ijcai.2026/494","url":"https://ijcai.org/proceedings/2026/494","pdf":"https://ijcai.org/proceedings/2026/0494.pdf","markdown":"https://ijcai.org/proceedings/2026/494.md","bibtex":"https://ijcai.org/proceedings/2026/bibtex/494","volume":{"slug":"2026","title":"Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2026","url":"https://ijcai.org/proceedings/2026/","year":2026,"location":"Bremen, Germany","dates":"15-21 August 2026","editors":"Diego Calvanese","isbn":"978-1-956792-09-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2026"},"section":"Machine Learning","track":"Main Track","number":494,"pages":{"first":4439,"last":4447},"keywords":["Machine Learning: Federated learning"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}