Table 4 The evaluation metrics for the generative model were obtained by executing the training pipeline ten times. Each iteration involved training the model and then utilizing it to generate synthetic data. Following the generation of data in each run, quality metrics for the produced synthetic data were calculated. The mean of these quality metrics was then computed to provide a comprehensive assessment of the synthesizer’s performance

From: A novel deep synthesis-based insider intrusion detection (DS-IID) model for malicious insiders and AI-generated threats