Reducing Instability in Synthetic Data Evaluation with a Super-Metric in MalDataGen

2511.16373v1 cs.AI, cs.LG 2025-11-22
Авторы:

Anna Luiza Gomes da Silva, Diego Kreutz, Angelo Diniz, Rodrigo Mansilha, Celso Nobre da Fonseca

Abstract

Evaluating the quality of synthetic data remains a persistent challenge in the Android malware domain due to instability and the lack of standardization among existing metrics. This work integrates into MalDataGen a Super-Metric that aggregates eight metrics across four fidelity dimensions, producing a single weighted score. Experiments involving ten generative models and five balanced datasets demonstrate that the Super-Metric is more stable and consistent than traditional metrics, exhibiting stronger correlations with the actual performance of classifiers.

Ссылки и действия