Research Of Malicious Android Application Detection Based On Static Multi-Grained Cascade Forest

With the continuous progress of mobile communication technologies and the impact of mobile network speed-up and fee-reduction policies,mobile smart terminals such as smart phones and tablet computers have been widely used in daily life and business environments.At the same time,a variety of malicious Android applications are constantly emerging and spreading.To protect Android users' data security and property safety,we urgently need effective detection tools,detection methods,and means to deal with more and more complicated security challenge.After learning the existing malicious android application detection methods and multi-grained cascade forest theory and learning from the experience of machine learning in other fields,this paper proposes a malicious android application detection model system based on static multi-grained cascade forest.The APK database module,APK feature extraction module,GCForest training module and unknown property APK detection module in this model system are introduced in detail.The core module of the model are APK feature extraction module and GCForest training module which extract feature data through decompilation techniques and use multi-grained scanning to perform feature conversion and supervised learning through cascade forests.Finally,this paper uses the published android application sample to carry out simulation experiments on the proposed model.In the experiment,we discussed the influences of the number of trees in the random forest when multi-grained scanning,the size of the sliding window,the sliding step length,and the number of random forest in the cascade layer,the number of trees in a single random forest in the cascade layer to the classification accuracy and the time consumed.Furthermore,we compared GCForest with SVM and RandomForest when the hyperparameter is set to optimal.The experiments show that the classification accuracy and AUC value of malicious android application detection model based on static multi-grained cascade forest are higher.