Privacy-Preserving Data Quality Evaluation in Federated Learning Using Influence Approximation
Primary Area: general machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: federated learning, client selection, vit, visual image transformers, image classification, transformers, data valuation, privacy preserving
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
TL;DR: Privacy-preserving data quality valuation technique for federated learning.
Abstract: In Federated Learning, it is crucial to handle low-quality, corrupted, or malicious data, but traditional data valuation methods are not suitable due to privacy concerns. To address this, we propose a simple yet effective approach that utilizes a new influence approximation called "lazy influence" to filter and score data while preserving privacy. To do this, each participant uses their own data to estimate the influence of another participant's batch and sends a differentially private obfuscated score to the Center of the federation. Our method has been shown to successfully filter out corrupted data in various applications, achieving a recall rate of over $>90\%$ (sometimes up to $100\%$) while maintaining strong differential privacy guarantees with epsilon values of less than or equal to one.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 3464
Loading