Mimic Android Malware Dynamic Detection Method Based On Enhanced Deep Learning

With the rapid development of mobile network,mobile devices have gradually become the preferred network terminal devices.At the same time,Android operating system has gradually become the preferred mobile operating system in the Chinese market.A large number of applications of Android operating system make more illegal people try to attack Android devices to obtain illegal benefits,which also leads to the proliferation of Android malware.In recent years,the escape technology of Android malware is gradually upgrading,and the difficulty of Android malware detection is increasing.In the face of the current severe situation of mobile network security,how to detect Android malware efficiently and accurately is an important topic.Firstly,a mimic Android malware dynamic detection structure based on enhanced deep learning is proposed.Using the idea of dissimilar redundancy structure in mimic security,input data randomly selects three heterogeneous redundancies in the architecture,namely enhanced LSTM model,enhanced GRU model and enhanced Capsule network model for training and prediction.Heterogeneous redundancies refer to different algorithms that implement the same function,which together constitute a set of functionally equivalent redundant isomers.The mimic Android malware dynamic detection structure based on enhanced deep learning can effectively detect Android malware and ensure the security of the detection structure itself.Then,based on the mimicry architecture,a dynamic detection method of Android malware based on enhanced deep learning is proposed.The enhanced LSTM model,enhanced GRU model and enhanced Capsule network model are proposed.Enhanced LSTM model,enhanced GRU model and enhanced Capsule network model can learn the historical information features of input data more effectively.In conclusion,this paper proposes a mimic Android malware dynamic detection method based on enhanced deep learning.The API call sequence of Android software is used as the input feature,and the mimicry architecture and enhanced deep neural network are used to detect Android malware dynamically.After several groups of comparative experiments,the results show that the mimic Android malware dynamic detection model based on enhanced deep learning has better detection ability than the traditional deep learning model,and has better defense performance than the traditional deep learning model.