New method enhances data utility under Local Differential Privacy

Researchers have developed a novel method to improve the utility of data collected under Local Differential Privacy (LDP). This approach selectively reduces noise in subspaces of the data that are most relevant to a specific task, rather than applying uniform noise across all dimensions. By identifying these critical subspaces using the Jacobian of a public model, the method reshapes the noise distribution from isotropic to anisotropic, enhancing data utility while maintaining privacy guarantees. Experiments show significant accuracy improvements on image classification tasks. AI

IMPACT This research could lead to more practical applications of privacy-preserving data collection in machine learning by improving the accuracy of models trained on such data.

RANK_REASON The cluster contains an academic paper detailing a new method for enhancing data utility under Local Differential Privacy. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• CIFAR-10-C
• Jacobian matrix
• Laplace
• Local Differential Privacy
• PrivUnit
• Youngmok Ha

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →