Research On Android Application Network Behavior Analysis Technology

With the rapid development of mobile internet,the Android platform has been widely used due to its openness and ease of use.The diversity and extensiveness of Android applications have also become an important factor affecting network security and management.Therefore,the network behavior analysis of Android applications has become one of the hotspots of current mobile internet research.However,due to the widespread use of cloud computing and CDN(Content Delivery Network)services,any identification of IP addresses or domain names in mobile network traffic is invalid,resulting in traditional traffic identification methods based on predefined ports and statistical features no longer applicable to mobile networks.At the same time,many Android applications are based on the HTTPS protocol,making it impossible to identify encrypted traffic using traditional machine learning methods.In view of the above problems,based on in-depth study of Android application network behavior characteristics,this thesis proposes a method of Android application network behavior classification based on deep learning.This method converts Android application network behavior classification problem into image classification problem,and then uses deep learning model to realize the classification of Android application network behavior.Aiming at the large amount of training data generation problems required by deep learning methods,this thesis proposes a method for generating Android application network behavior test cases.Using software static analysis technology to analyze and characterize the network-related behavior context of Android applications,obtaining traffic data about application network behavior more accurately and efficiently.The main work of the thesis is as follows:(1)Proposing a method for generating Android application network behavior test cases.Aiming at the large amount of Android application network traffic data required by the deep learning model,analyzing the network-related execution path can more accurately obtain context information related to the application network behavior and reduce the analysis of nonrelated information.Using the backward program slicing and symbol execution technology,starting with network-related system calls,constructing the Android application network behavior context related to network traffic generation,and generating the test cases with good network behavior coverage accordingly.On this basis,Combined with automated testing tools,obtaining network data traffic that accurately characterizes the behavior of Android applications effectively.(2)Proposing a method based on deep learning for Android application network behavior analysis.Applying the deep learning method to the classification problem of Android application network behavior.By transforming the application network behavior classification problem into image classification problem,using the convolutional neural network model to self-learn the network behavior characteristics in the image to realize the Android application network behavior classification.(3)Designing and implementing an Android application network traffic classification system based on network behavior analysis.Android application network traffic classification system using test case generation method and deep learning method to claasify network traffic.Android application network traffic classification system can effectively identify Android application network traffic,with high identify accuracy and efficiency.