"Code, Dataset and Zero-Day Evaluation Framework for Calibrated Family-Agnostic DGA Detection with Feature-Aware Dual-Input DL"

"This repository contains the full implementation, dataset splits, feature extraction pipeline and experimental evaluation framework used in the paper:\u201cCalibrated and Family-Agnostic Domain Generation Algorithm Detection with Robust Zero-Day Generalization via Feature-Aware Dual-Input Deep Learning.\u201dThe repository includes:\u2022 Source code for the dual-input CNN\u2013BiLSTM\u2013Attention model.\u2022 Feature-aware tokenization and numeric feature extraction modules.\u2022 Dataset partitions used for training and evaluation.\u2022 Leave-One-Out (LOO) family-wise zero-day evaluation setup.\u2022 Scripts for computing ROC-AUC, Precision, Recall, F1-score and False Positive Rate.\u2022 Experimental results and logs for reproducibility.This repository is intended to support reproducible research in DGA detection, DNS security and zero-day malware domain detection."