This improvement comes from the risk of constructing neural networks with the next number of doubtlessly numerous layers and is named Deep Learning. In this paper, we try to switch these performance improvements to mannequin the malware system name sequences for the aim of malware classification. We assemble a neural network based mostly on convolutional and recurrent community layers so as to obtain one of the best features for classification. This method we get a hierarchical characteristic extraction architecture that mixes convolution of n-grams with full sequential modeling.