Research And Implementation Of Malicious Application Detection Technology Based On Deep Learning

Android and Windows are currently the most popular mobile operating system and desktop operating system respectively.Their software installation and management methods are relatively open and loose,and there are a large number of malicious applications on these two operating systems.Various types of malicious applications bring privacy disclosure,business secret disclosure,public information disclosure and economic losses to individuals,businesses and government departments.In this thesis,the problem of malicious application detection under two systems is studied.A malicious application detection model combined with one-dimensional convolutional neural network and recurrent neural network and its supporting preprocessing methods are proposed.Based on the microservice architecture concept,a distributed malicious application detection engine is designed and implemented.This thesis proposes a deep learning based malicious application detection technology,including a detection model and its associated preprocessing method.The detection technology divides the information extracted from the application files into two categories,discrete information and sequence information.Two branch models based on one-dimensional convolutional neural network and recurrent neural network are designed respectively.These two branch models merge to form the final detection model.The preprocessing method designs the fast mode and the slow mode according to the degree of processing of the application file.The detection technology integrates the detection problems under the two operating systems of Android and Windows into one framework and improves the accuracy of 2?3%in the slow preprocessing mode compared with existing methods.In the fast preprocessing mode,compared with the existing detection technology,the accuracy is reduced by less than 2%,and the detection speed is increased by more than 5 times.Based on the microservice architecture concept,this thesis designs and implements a distributed malicious application detection engine system.The detection engine is supported by the container technology Docker and the container orchestration technology Kubernetes.The interface service,cache service,preprocessing service,model discrimination service,message queue service,data access service,and configuration management service are designed.These service modules cooperate to form a detection engine system,which has high availability and scalability and can cope with large-scale detection requirements.