CausShield: Sample Reconstruction-Resilient Vertical FL via Causal Representation Learning
Researchers have developed CausShield, a new method to enhance privacy in vertical federated learning (VFL). This approach uses causal representation learning to distinguish between task-relevant and task-irrelevant features within data. By separating these components, CausShield aims to protect sensitive private information while maintaining model utility, offering a more robust defense against sample reconstruction attacks than existing methods. AI
IMPACT Enhances privacy guarantees for distributed machine learning systems, potentially enabling more sensitive data collaborations.