Learning to Generate Image Embeddings with User-level Differential Privacy
This addresses privacy concerns for users in applications like facial recognition and species classification by enabling secure, large-scale model training with minimal performance loss.
The paper tackled the challenge of training large image-to-embedding models with user-level differential privacy, which existing methods fail for due to large class spaces, and proposed DP-FedEmb, achieving strong privacy guarantees (ε<4) with utility drops within 5% on benchmark datasets like DigiFace and iNaturalist.
Small on-device models have been successfully trained with user-level differential privacy (DP) for next word prediction and image classification tasks in the past. However, existing methods can fail when directly applied to learn embedding models using supervised training data with a large class space. To achieve user-level DP for large image-to-embedding feature extractors, we propose DP-FedEmb, a variant of federated learning algorithms with per-user sensitivity control and noise addition, to train from user-partitioned data centralized in the datacenter. DP-FedEmb combines virtual clients, partial aggregation, private local fine-tuning, and public pretraining to achieve strong privacy utility trade-offs. We apply DP-FedEmb to train image embedding models for faces, landmarks and natural species, and demonstrate its superior utility under same privacy budget on benchmark datasets DigiFace, EMNIST, GLD and iNaturalist. We further illustrate it is possible to achieve strong user-level DP guarantees of $ε<4$ while controlling the utility drop within 5%, when millions of users can participate in training.