{"ID":2845934,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2511.03806","arxiv_id":"2511.03806","title":"FusionDP: Foundation Model-Assisted Differentially Private Learning for Partially Sensitive Features","abstract":"Ensuring the privacy of sensitive training data is crucial in privacy-preserving machine learning. However, in practical scenarios, privacy protection may be required for only a subset of features. For instance, in ICU data, demographic attributes like age and gender pose higher privacy risks due to their re-identification potential, whereas raw lab results are generally less sensitive. Traditional DP-SGD enforces privacy protection on all features in one sample, leading to excessive noise injection and significant utility degradation. We propose FusionDP, a two-step framework that enhances model utility under feature-level differential privacy. First, FusionDP leverages large foundation models to impute sensitive features given non-sensitive features, treating them as external priors that provide high-quality estimates of sensitive attributes without accessing the true values during model training. Second, we introduce a modified DP-SGD algorithm that trains models on both original and imputed features while formally preserving the privacy of the original sensitive features. We evaluate FusionDP on two modalities: a sepsis prediction task on tabular data from PhysioNet and a clinical note classification task from MIMIC-III. By comparing against privacy-preserving baselines, our results show that FusionDP significantly improves model performance while maintaining rigorous feature-level privacy, demonstrating the potential of foundation model-driven imputation to enhance the privacy-utility trade-off for various modalities.","short_abstract":"Ensuring the privacy of sensitive training data is crucial in privacy-preserving machine learning. However, in practical scenarios, privacy protection may be required for only a subset of features. For instance, in ICU data, demographic attributes like age and gender pose higher privacy risks due to their re-identifica...","url_abs":"https://arxiv.org/abs/2511.03806","url_pdf":"https://arxiv.org/pdf/2511.03806v1","authors":"[\"Linghui Zeng\",\"Ruixuan Liu\",\"Atiquer Rahman Sarkar\",\"Xiaoqian Jiang\",\"Joyce C. Ho\",\"Li Xiong\"]","published":"2025-11-05T19:13:10Z","proceeding":"cs.LG","tasks":"[\"cs.LG\"]","methods":"[]","has_code":false}