Research Of Malicious Code Detection Based On Deep Learning

At present,the network security problem caused by malicious code attack has become an important factor affecting the development of the Internet.Malicious code will attack and steal core data and sensitive information,even directly damage the computer system and network,which is one of the major threats in the field of network security.In this paper,a quantitative analysis of the malicious code detection literature is carried out,and introduce deep learning ideas,and a hierarchical detection model based on cascading and Deep Belief Network is constructed and implemented.The main contents of this paper are as follows:1.Aiming at the problems of low accuracy of traditional malicious code detection models and complex feature extraction,deep learning ideas are introduced,through deep belief network models to automatically learn sample features and train detection models.The image texture features and instruction frequency features of the malicious code are selected respectively,and the two features are fused to train the deep belief network detection model to improve the detection accuracy.2.In order to further improve the detection accuracy,a hierarchical detection model of malicious code based on cascading and deep belief networks is proposed.The cascade structure performs the first layer detection by matching Simhash signature values,LBP cascade sparse representation,and matching disassembly features;then the image texture features and instruction frequency features of the extracted malicious code are fused to train the Deep Belief Network,and the Deep Belief Network is used for deep detection.The model proposed in this paper can be tested from simple to complex,which improves the degree of automation and accuracy of test and scalability.3.The feasibility of the model in this paper is verified through experiments.Under the selection of image texture features,instruction frequency features and fusion features of the two,compared and analyzed the results through different detection models,it is found that the detection model based on cascade and Deep Belief Network proposed in this paper can obtain higher accuracy than other models.To further improve the applicability of the model,a malicious code detection system is designed and implemented.