Postprocessing for Iterative Differentially Private Algorithms
This work addresses accuracy limitations in privacy-preserving machine learning for data analysts, but it appears incremental as it builds on existing iterative methods.
The paper tackles the problem of improving accuracy in iterative differentially private algorithms by proposing a post-processing method that utilizes intermediate outputs, which contain data knowledge, to enhance the final result.
Iterative algorithms for differential privacy run for a fixed number of iterations, where each iteration learns some information from data and produces an intermediate output. However, the algorithm only releases the output of the last iteration, and from which the accuracy of algorithm is judged. In this paper, we propose a post-processing algorithm that seeks to improve the accuracy by incorporating the knowledge on the data contained in intermediate outputs.