Researchers have developed a new framework that adapts Shapley-value-based feature attribution to address data privacy concerns. This approach captures both disclosure risk and data utility at the feature level, offering a more granular perspective than existing dataset-level methods. The framework is designed to be agnostic to various data masking, statistical, and machine learning techniques, demonstrating effectiveness in reducing disclosure risk while maintaining data utility in experimental results. AI
IMPACT This research could lead to more robust data masking techniques, improving privacy in AI model training and deployment.
RANK_REASON This is a research paper detailing a novel framework for data privacy using Shapley values. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
