(НИЯУ МИФИ, 2025) ANAEDEVHA ,R. N.; TROFIMOV, A. G.; Трофимов, Александр Геннадьевич
We propose a dual-branch architecture for cross-domain intrusion detection systems across six datasets. The cloud branch employs game-theoretic federated learning (FL) with Byzantine robustness and differential privacy (DP) guarantees (Edge-IIoT, Container, SOC), achieving 95.7–96.9% accuracy. The system maintains ε-DP, Byzantine resilience, and adversarial robustness via progressive adversarial robust distillation (PARD) and Probably Approximately Correct (PAC)-Bayesian regularization.