Research on unsupervised feature learning for Android malware detection based on Restricted Boltzmann Machines

摘要: Abstract Android malware detection has attracted much attention in recent years. Existing methods mainly research on extracting static or dynamic features from mobile apps and build model by machine learning algorithms. The number of extracted maybe high. As a result, the data suffers high dimensionality. In addition, to avoid being detected, is varied hard obtain first place. To detect zeroday malware, unsupervised were applied. such case, feature reduction method an available choice reduce this paper, we propose algorithm called Subspace based Restricted Boltzmann Machines (SRBM) for reducing dimensionality detection. Multiple subspaces original are firstly searched. And then, RBM built each subspace. All outputs hidden layers trained RBMs combined represent lower dimension. experimental results OmniDroid, CIC2019 CIC2020 datasets show that learned SRBM perform better than ones other when performance evaluated clustering evaluation metrics, i.e., NMI, ACC Fscore.