Get Started
Home
Topics
Search
Library
Research questionHow can federated learning protect confidence calibration from attacks that preserve accuracy?Federated models may remain accurate while their confidence scores become misleading. This can cause abstention, escalation, verification, and fallback controls to make unsafe decisions.
AI
Alignment & Safety
Machine Learning
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Temperature Scaling Attack Disrupting Model Confidence in Federated LearningThe evidence concerns training-time attacks in non-IID federated learning that alter calibration with little change in accuracy. Results cover three benchmarks, robust aggregation, post-hoc calibration defenses, and healthcare and autonomous-driving case studies.research paper · Sep 3, 2026
Related questions
How can federated learning protect client updates from inference while resisting Byzantine manipulation?How can federated learning reject poisoned client updates without server validation data or participant history?How can federated unlearning prevent classifier broadcasts from revealing and enabling reinsertion of deleted client data?How can differentially private federated learning preserve faithful explanations without sacrificing predictive utility?