| Title: |
Differential Confounding Privacy and Inverse Composition |
| Authors: |
Zhang, Tao, Malin, Bradley A., Raviv, Netanel, Vorobeychik, Yevgeniy |
| Publication Year: |
2024 |
| Collection: |
Computer Science |
| Subject Terms: |
Computer Science - Cryptography and Security |
| More Details: |
Differential privacy (DP) has become the gold standard for privacy-preserving data analysis, but its applicability can be limited in scenarios involving complex dependencies between sensitive information and datasets. To address this, we introduce Differential Confounding Privacy (DCP), a framework that generalizes DP by accounting for broader causal relationships between secrets and datasets. DCP adopts the $(\epsilon, \delta)$-privacy framework to quantify privacy loss, particularly under the composition of multiple mechanisms accessing the same dataset. We show that while DCP mechanisms retain privacy guarantees under composition, they lack the graceful compositional properties of DP. To overcome this, we propose an Inverse Composition (IC) framework, where a leader-follower model optimally designs a privacy strategy to achieve target guarantees without relying on worst-case privacy proofs. Experimental results validate IC's effectiveness in managing privacy budgets and ensuring rigorous privacy guarantees under composition. |
| Document Type: |
Working Paper |
| Access URL: |
http://arxiv.org/abs/2408.12010 |
| Accession Number: |
edsarx.2408.12010 |
| Database: |
arXiv |
