Research on a Malicious Code Detection Method Based on Convolutional Neural Network in a Domestic Sandbox Environment

Abstract

AbstractFor malicious code detection, the paper proposes an improved serialization detection method based on convolutional neural network algorithm, it adopts the architecture of “domestic environment virtual sandbox + convolutional neural network detection model + dynamic simulation”. First, extract the features of the API sequence, use the Densenet model to detect on the basis of redundant information preprocessing, and then use the characteristics of the convolutional neural network in deep learning to process time series data to directly model and learn the sequence. Finally, based on virtualization technology, a simulation experiment is carried out in the virtual sandbox environment of a domestic safe and reliable operating system. Through three comparative experiments of malicious code detection accuracy, missed detection rate and efficiency, The results show that the improved method has high efficiency and accuracy in detecting a large number of malicious codes, and it can be applied to the detection of malicious codes in a safe and controllable operating system.KeywordsIndustrial information securityVirtualizationSandboxMalicious codeConvolutional neural network