Publications

Detecting Zero-Day Attacks via Reconstruction of Feature Influence and Model Uncertainty

IEEE Transactions on Network and Service Management (TNSM), 2026

Combining XAI-based feature-influence reconstruction with ensemble uncertainty to recognize zero-day attacks that a multiclass NIDS classifier would otherwise mislabel as known threats.

A Unified Framework for Trustworthy Machine Learning for Intrusion Detection Systems

IFIP Networking Conference (IFIP Networking), Doctoral Consortium, 2026

Energy Cost of Enhancing Reliability of Machine Learning Models for Edge IoT Security

Wireless On-demand Network Systems and Services Conference (WONS), 2026

Improving reliability in ML-based IoT security comes at an energy cost. This work quantifies and analyzes that trade-off for edge deployments.

Towards Better-Calibrated ML Models for Reliable Network Intrusion Detection via Calibration-Aware SHAP-Based Feature Selection

IEEE International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob), 2025

Calibration-aware SHAP-based feature selection improves reliability and confidence estimation for ML-based intrusion detection systems.

Reducing Complexity and Enhancing Predictive Power of ML-based Lightpath Quality of Transmission Estimation via SHAP-Assisted Feature Selection

International Conference on Optical Network Design and Modeling (ONDM), 2024

SHAP-assisted feature selection reduces model complexity while improving prediction accuracy for optical network QoT estimation.