Get Started
Home
Topics
Search
Library
Research questionHow do privacy guarantees and Byzantine robustness affect rare-attack coverage in imbalanced federated intrusion detection?Federated intrusion detectors must learn from highly imbalanced traffic while client updates are privacy-protected and may be malicious. Privacy noise can obscure minority-class signals, while robust aggregation may further suppress them, making overall accuracy an incomplete measure of detection quality.
AI
Machine Learning
Research Paper
Statistical Machine Learning
Technology
Latest papersRecent research connected to this question, newest first.Privacy, Robustness, and Fairness Trade-offs in Federated Intrusion Detection: Geometric Indistinguishability at the Aggregation InterfaceThe evidence is limited to a UNSW-NB15 case study using DP-SGD with coordinate-wise median aggregation under label-flip and model-poisoning attacks. The study reports initial evidence that strong privacy can worsen rare-attack detection; some degradation may reflect training miscalibration, while ultra-rare categories may retain a performance floor after epsilon-dependent tuning.research paper · Sep 3, 2026
Related questions
How can federated learning protect client updates from inference while resisting Byzantine manipulation?How can federated learning protect confidence calibration from attacks that preserve accuracy?How can federated learning stay reliable when heterogeneous edge clients are intermittently available and availability changes over time?How can federated LiNGAM causal discovery remain reliable under near-symmetric noise without centralizing data?