Google AI and Tel Aviv Researchers Introduce FriendlyCore: A Machine Learning Framework For Computing Differentially Private Aggregations - MarkTechPost

Data analysis revolves around the central goal of aggregating metrics. The aggregation should be conducted in secret when the data points match personally identifiable information, such as the records or activities of specific users. Differential privacy (DP) is a method that restricts each data point's impact on the conclusion of the computation. Hence it has become the most frequently acknowledged approach to individual privacy. Although differentially private algorithms are theoretically possible, they are typically less efficient and accurate in practice than their non-private counterparts.