Doppio : Communication-Efficient and Secure Multi-Party Shuffle Differential Privacy

Auditing and Mitigating Privacy Leakage in Cloud-Edge Collaborative Decoding

CoVeil is proposed, a defense mechanism which dynamically optimizes transmitted signals to suppress leakage during decoding time while preserving the collaborative quality, and consistently improves the privacy-utility trade-off over existing baselines by reducing data leakage.

Private and Verifiable Outsourcing of Open-Weight LLM Inference

An End-to-End System for Efficient Multiparty Computation in Privacy-Preserving Machine Learning

The proposed Secure Multiparty Computation protocol enables collaborative training of linear and logistic regression models while providing formal privacy guarantees for participant data, and adapts the iterative gradient descent algorithm to operate securely over secretly shared vectors.

Privacy-preserving secure data sharing in edge-cloud collaborative environments

A privacy-preserving, secure data-sharing framework tailored for edge-cloud collaborative architectures that minimizes the computational overhead on the terminal side while safeguarding user privacy, and effectively reduces the overhead associated with user joining and revocation within the same group.

TrustCloud-FL: A Self-Healing and Privacy-Preserving Architecture for Secure Analytics Across Multi-Tenant Hybrid Clouds

Cloud platforms let organizations share computing capacity, data services, and machine-learning tools across many business units, regions, and providers. The same flexibility also creates a difficult security problem: raw records may cross tenant boundaries, model updates may be poisoned, automated responses may interr...

On the Limits of Privacy in Federated Learning: Personalization and Adversarial Analysis