Quantum-Resilient Federated Cyber Defense Framework (Independent Research)
Conducted adversarial robustness testing, ablation studies, and scalability benchmarking for a federated edge-AI intrusion detection framework across multiple benchmark datasets. Used explainability signal disagreement (SHAP, LIME, and TabNet attention) to generate an evaluation target for adversarial manipulation detection. Assessed model performance under perturbation/noise conditions to validate detection reliability in adversarial settings. • Built Tri-ECS to score explanation consistency as an adversarial detection signal. • Implemented TAFA-v2 with meta-learned aggregation weights informed by explainability consistency. • Ran GPU-accelerated simulation experiments for multiple ML model families (RF, XGBoost, LSTM, Autoencoder, Isolation Forest). • Performed formal ablations and robustness tests across different federated node configurations.