#differential-privacy

The latest AI and developer news about #differential-privacy, with the original source and a short summary.

Google Applies Externally Verifiable Differential Privacy to Gboard Model Training via TEE-Based Federated Learning

Shortens Gboard training period from 2 months to 3 weeks and improves privacy budget by 3x

·3·2026.10.11 13:00

Google Research Introduces TEE-Based Federated Learning System for Provably Private Training

Gboard has already adopted the new system, achieving faster training speeds and stronger privacy guarantees for next-word prediction models.

·2·2026.10.02 09:00

DP-FedSOFIM: Second-Order-Based Federated Learning Optimization

It proposes a federated learning optimization technique that applies second-order computation without increasing clients' communication load or privacy cost.

·2026.07.28 15:00

Behavioral Privacy Leakage in Agent Negotiations: Formalizing and Mitigating Inference Attacks via Randomized Policies

To prevent attacks that infer personal information through the behavioral patterns of negotiation agents, researchers proposed a new negotiation policy applying differential privacy.

·2·2026.07.10 09:00

Differentially Private Clustered Federated Learning with Privacy-Preserving Initialization and Normality-Based Aggregation

PINA improves the performance of differentially private clustered FL through LoRA sketching and normality-based aggregation.

·2·2026.05.07 00:00

Preserving Privacy in AI Training Data

Sensitive AI training data proved vulnerable to membership inference and reconstruction attacks, with DP and MPC serving as defenses.

·1·2026.04.30 02:00