DP$^2$-VAE: Differentially Private Pre-trained Variational Autoencoders
This work addresses privacy concerns in machine learning for sensitive data like medical records and face images, offering a flexible solution that enhances utility in differentially private generative models, though it is incremental as it builds on existing VAE frameworks.
The paper tackles the challenge of balancing utility and privacy in differentially private generative models by proposing DP^2-VAE, a training mechanism for variational autoencoders that uses pre-training on private data to minimize perturbation noise, resulting in improved utility under the same privacy constraints, with empirical superiority demonstrated on image datasets across various privacy budgets and metrics.
Modern machine learning systems achieve great success when trained on large datasets. However, these datasets usually contain sensitive information (e.g. medical records, face images), leading to serious privacy concerns. Differentially private generative models (DPGMs) emerge as a solution to circumvent such privacy concerns by generating privatized sensitive data. Similar to other differentially private (DP) learners, the major challenge for DPGM is also how to achieve a subtle balance between utility and privacy. We propose DP$^2$-VAE, a novel training mechanism for variational autoencoders (VAE) with provable DP guarantees and improved utility via \emph{pre-training on private data}. Under the same DP constraints, DP$^2$-VAE minimizes the perturbation noise during training, and hence improves utility. DP$^2$-VAE is very flexible and easily amenable to many other VAE variants. Theoretically, we study the effect of pretraining on private data. Empirically, we conduct extensive experiments on image datasets to illustrate our superiority over baselines under various privacy budgets and evaluation metrics.