UDCDGA: A large-scale balanced dataset for Domain Generation Algorithm detection

UDCDGA is a large-scale dataset designed to support research on Domain Generation Algorithm (DGA) detection and analysis. The dataset was constructed using UDCDGA_Generator, a reproducible and extensible software framework for building DGA datasets from benign-domain sources and multiple DGA implementations. The dataset contains 8,450,556 unique domain names, equally distributed between 4,225,278 benign and 4,225,278 DGA-generated domains. The DGA class includes domains generated from multiple DGA implementations and generation mechanisms. Each domain is represented by a predefined set of 54 lexical, structural, statistical, and n-gram-based features suitable for machine-learning-based analysis. The primary dataset is distributed in Apache Parquet format. Complementary metadata describe the dataset schema, construction statistics, feature representation, execution parameters, and integrity information, supporting traceability and reproducibility. UDCDGA is intended for cybersecurity research, including DGA detection, machine-learning benchmarking, feature analysis, explainable artificial intelligence, and the evaluation of detection approaches against heterogeneous DGA implementations. The dataset was generated using UDCDGA_Generator v1.0.0.