Android Malware Detection Method Based On Improved Naive Bayessian And Permissions

With the rapid development of the information technology, it is more and more popular to mobile terminal devices. The Android system has been widely used, and more and more people has begun to concern about Android security issues. Permission management is the key of Android security mechanism, and there are many researchers who have studied Android security issues by analyzing the permission mechanism. Most researchers in the past just considered that the permission of the sample were malicious permission, and little considered that the new permission in the test sample might be malicious. In order to detect Android malware more effectively, an Android malware detection model was proposed based on improved naive bayes classification. The research contents are two parts as follows.Considering the unknown permission that may be malicious in detection samples, and in order to improve the Android detection rate, the algorithm of malware detection is proposed based on improved naive bayes. Considering the limited training samples, limited permissions, and the new malicious permissions in the test samples, we used the impact of the new malware permissions and training permissions as the weight. The weighted Naive Bayesian algorithm improves the Android malware detection efficiency.Taking into account the detection model, we proposed a detection model of permissions and information theory based on the improved naive bayes algorithm. We analyzed the correlation of the permission. By calculating the Pearson correlation coefficient, we determined the value of Pearson correlation coefficient r, and delete the permissions whose value r is less than the threshold ?and get the new permission set. Finally, we got the improved detection model by clustering based on imformation theory.We detected the 1725 Android malware and 945 non malicious application of multiple data sets in the same simulation environment. By analyzing the detection rate and false detection rate of malicious and non-malicious, the experiment results prove that it is efficiency and accuracy, and it performs better than other methods in efficiency and accuracy when detecting the latest malicious applications.