• Visual Utility of Differentially Private Scatterplots under US Census data

• Project Year: 2023
• REU Student (s):   Martha-Victoria Parizot | Harvey Mudd College CA  
• Student 1 Institution: Harvey Mudd College
• Project Mentor: Anand Sarwate
• Project Mentor Area: Electrical and Computer Engineering
• Project Abstract: In data-driven research fields like healthcare and technology, access to sensitive individual information is crucial for generating valuable insights. However, privacy regulations pose challenges for publishing such data, hindering researchers' access. Differential privacy has emerged as a promising solution, allowing researchers to learn from sensitive data while protecting individual privacy. This research focuses on generating differentially private scatterplots, a common data visualization tool, while retaining visual utility. The approach combines strategies of partitioning data, adding calibrated noise, and post-processing to suppress noise. The study explores factors like ε (privacy level), algorithms, bin size, data distribution, and sample size to understand their impact on visual utility. Two small datasets are used, derived from the US Census and World Health Organization, with direct application for Differential Privacy. The results highlight the trade-offs between privacy and visualization integrity, aiding researchers in making informed decisions to balance privacy and data visualization accuracy. Future directions include exploring more sophisticated point regeneration techniques and integrating differentially private heatmaps techniques to improve visual accuracy