Detecção de Malwares Android: uma análise ampla de datasets e reprodutibilidade

Abstract

In this work, we present an analysis and research results on datasets and the reproducibi-lity of prediction models in the context of Android malware detection. The development and results of the research are organized in the form of three technical-scientific articles, two of which are published in the VI Workshop Regional de Segurança da Informação e de Sistemas Computacionais (WRSeg 2021) and a third stage in an advanced stage of development. As main contributions we can highlight: (i) analysis and discussion of the datasets used by researches in detection of Android malware; (ii) mapping of the datasets availability; (iii) identification of incompleteness and inconsistencies in the works; (iv) cataloging and classification of 84 data sources; (v) identification of inconsistency related to information about the actuality of the datasets; (vi) good practices recommendations for research works that use machine learning equipment to detect Android malware; and (vii) obstacles identification related to works reproducibility.