Differential privacy mechanisms can protect tabular data in in-context learning, ensuring privacy while maintaining performance.
Proposing the Independent Component Laplace Process (ICLP) mechanism for differential privacy in functional summaries.
Proposing a novel k-stars LDP algorithm for (p, q)-clique enumeration with improved utility and privacy protection.
API-based AUG-PE algorithm generates high-quality DP synthetic text without model training, outperforming traditional DP finetuning methods.
Modeling privacy policies using the Privacy Policy Permission Model (PPPM) can help identify gaps, inconsistencies, and potential privacy risks in organizations' data handling practices.
Citizens and small data holders face challenges in exercising GDPR rights, requiring innovative solutions like privacy dashboards and services.
提案されたVPASプロトコルは、データプライバシーを保護しながら入力検証と公開検証を実現することで、集計統計の効率的な処理を可能にします。
Generating differentially private synthetic data using foundation model APIs without training.
The author examines the flaws in measuring data anonymity vulnerabilities, highlighting errors in statistical inference baselines and base rate assumptions.
The author explores the use of differential privacy to protect tabular data in the context of in-context learning, proposing two frameworks - LDP-TabICL and GDP-TabICL - that offer privacy guarantees while maintaining performance.