Differentially Private Genome Data Dissemination Through Top-Down Specialization

Explore a 24-minute conference talk from USENIX HealthTech '14 that presents a novel approach for disseminating genomic data while maintaining differential privacy. Learn about an algorithm that splits raw genome sequences into blocks, subdivides them using a top-down method, and adds noise to counts for privacy protection. Discover how this technique can potentially retain higher data utility compared to baseline methods for a given privacy budget. Understand its applicability to heterogeneous data, including combined medical and genomic records. Gain insights into possible future improvements, such as refining sequence splitting heuristics and introducing scoring functions in the data generalization process.