Behavioral and Psychometric Feature Matrix for Insider Threat Detection (Derived from CERT r4.2)

This dataset provides a consolidated and engineered feature matrix designed for Insider Threat detection using Unsupervised Machine Learning and Deep Learning architectures. It is derived from the synthetic CERT r4.2 dataset (Software Engineering Institute, Carnegie Mellon University). The dataset bridges traditional User and Entity Behavior Analytics (UEBA) with psychometric profiling, based on the Critical Path to Insider Risk (CPIR) model. It contains over 330,000 user-day records, fusing technical volumetric data (logons, HTTP, USB, file activity, after-hours ratios) with static personality traits (Big Five) and dynamic affective drift metrics (Z-Scores derived from NLP sentiment analysis of organizational emails). Two files are included: master_behavioral_matrix.parquet: The aggregated base matrix. feature_matrix_v1.parquet: The final engineered dataset with rolling windows and normalized behavioral features, ready for ML/DL modeling.