Android Malware Detection Based On Feature Weighting With Joint Optimization Of Parameters

The eco-friendly and open-source nature of Android OS has made it the worlds number one mobile OS,however,its open-source nature has also led to the rapid growth of Android malware,which poses a huge threat to mobile devices and users;therefore,Android malware detection is an important problem that needs to be urgently researched and solved.There are basically two problems in the research work on Android malware detection,one is that the difference in importance between features is ignored,i.e.,the importance of each feature and the threat to the Android system is different;the other problem is that the classifiers selected for detection are insensitive to weights,which further reduces the importance of feature weights.Therefore,to address the above two problems,this paper designs an Android malware detection scheme JOWMDroid based on joint feature weighting and parameter optimization.The scheme avoids putting all features into a set to train the classifier for the final detection and assigns different weights to each feature according to the difference in importance of the features.First,eight classes of features are extracted from the Android application package,and then feature selection is performed for these eight classes using an information gain algorithm to select a certain number of the most important features.Second,an initial weight is calculated for each selected feature by training three machine learning models,and then the initial weight of each feature is mapped to the final weights using five selfdesigned parameterized weight mapping functions.Finally,the parameters of the weight mapping function and different classifiers are jointly optimized using a differential evolutionary algorithm.The experimental results show that this scheme outperforms four commonly used feature weighting methods at present and improves the competitiveness of weight-aware classifiers.