Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Defogger: A Visual Analysis Approach for Data Exploration of Sensitive Data Protected by Differential Privacy

Xumeng Wang, Shuangcheng Jiao, Chris Bryan

TL;DR

The paper addresses the challenge of exploring datasets protected by differential privacy, where limited privacy budgets and noisy responses hinder typical visual exploratory workflows. It introduces Defogger, a visual-analysis tool that uses a reinforcement-learning recommender to suggest exploration strategies and employs a Mosaic-based uncertainty visualization to communicate DP-induced uncertainty. Through two case studies and a user study, the work demonstrates that Defogger improves privacy-budget efficiency while aligning exploration with user intent and providing interpretable uncertainty guidance. The contributions span a novel DP-aware exploration pipeline, an RL-based strategy generator, and an integrated interface that supports information reservation, data-request declaration, and uncertainty visualization, with practical implications for privacy-preserving data analysis and visualization design.

Abstract

Differential privacy ensures the security of individual privacy but poses challenges to data exploration processes because the limited privacy budget incapacitates the flexibility of exploration and the noisy feedback of data requests leads to confusing uncertainty. In this study, we take the lead in describing corresponding exploration scenarios, including underlying requirements and available exploration strategies. To facilitate practical applications, we propose a visual analysis approach to the formulation of exploration strategies. Our approach applies a reinforcement learning model to provide diverse suggestions for exploration strategies according to the exploration intent of users. A novel visual design for representing uncertainty in correlation patterns is integrated into our prototype system to support the proposed approach. Finally, we implemented a user study and two case studies. The results of these studies verified that our approach can help develop strategies that satisfy the exploration intent of users.

Defogger: A Visual Analysis Approach for Data Exploration of Sensitive Data Protected by Differential Privacy

TL;DR

The paper addresses the challenge of exploring datasets protected by differential privacy, where limited privacy budgets and noisy responses hinder typical visual exploratory workflows. It introduces Defogger, a visual-analysis tool that uses a reinforcement-learning recommender to suggest exploration strategies and employs a Mosaic-based uncertainty visualization to communicate DP-induced uncertainty. Through two case studies and a user study, the work demonstrates that Defogger improves privacy-budget efficiency while aligning exploration with user intent and providing interpretable uncertainty guidance. The contributions span a novel DP-aware exploration pipeline, an RL-based strategy generator, and an integrated interface that supports information reservation, data-request declaration, and uncertainty visualization, with practical implications for privacy-preserving data analysis and visualization design.

Abstract

Differential privacy ensures the security of individual privacy but poses challenges to data exploration processes because the limited privacy budget incapacitates the flexibility of exploration and the noisy feedback of data requests leads to confusing uncertainty. In this study, we take the lead in describing corresponding exploration scenarios, including underlying requirements and available exploration strategies. To facilitate practical applications, we propose a visual analysis approach to the formulation of exploration strategies. Our approach applies a reinforcement learning model to provide diverse suggestions for exploration strategies according to the exploration intent of users. A novel visual design for representing uncertainty in correlation patterns is integrated into our prototype system to support the proposed approach. Finally, we implemented a user study and two case studies. The results of these studies verified that our approach can help develop strategies that satisfy the exploration intent of users.
Paper Structure (23 sections, 2 equations, 7 figures, 2 tables)

This paper contains 23 sections, 2 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Processes of Exploratory Visual Analysis
  4. Practices in Differentially Private Publication
  5. Visualizations for Uncertainty Representation
  6. Scenario Overview and Requirements Analysis
  7. Practice under Privacy Limitations
  8. Use Case
  9. Requirement Analysis
  10. Approach
  11. Pipeline
  12. Recommendation of Exploration Strategies
  13. Interface Design
  14. Information Reservation
  15. Data Request Declaration
  16. ...and 8 more sections

Figures (7)

  • Figure 1: Set divisions used in a data request.
  • Figure 2: The pipeline for visual exploration with the restriction of DP, which mainly consists of a recommendation model and three iterative modules. In the information reservation module, we take the use case in Sect. \ref{['sec:uc']} as an example.
  • Figure 3: Visualizations for uncertainty illustration. (a) Histogram with error bars for uncertainty in distribution over a single attribute. (b) Heatmap matrix representing correlations among multiple attributes, which requires grid-based uncertainty representation. (c) Two alternatives for uncertainty representation used in heatmap matrices.
  • Figure 4: Visualizations used in the first data request.
  • Figure 5: Visualizations used in the second data request, consisting of (a) the intent graph specified by the user, (b) details of request declaration, and (c) distribution comparison between policy: B clients and all clients.
  • ...and 2 more figures