Android Malware Detection Based On Improved Naive Bayesian Algorithm

Now, smart phones which based on Android platform accounts for the lion’s share of the smart phone market, the number of malicious software on the Android platform shows increasing growth trend, study on Android malware has become a hot topic of mobile security. The application of machine learning techniques to Android malware detection is very common, using Naive Bayesian classification algorithm for Android malware detection more, but in the existing Android malware based on Naive Bayesian classification algorithm research the shortcomings of Naive Bayesian classification algorithm has not been considered, such as the lack of mutual independence between the characteristic attributes of hypothetical conditions and the heavy weight of each feature attribute values are treated the same, thus affect the detection performance of Android malware.Considering these shortcomings mentioned above, in this paper we propose the following two improvements, The first point is the improvement of extracting Naive Bayesian classification algorithm for preprocessing feature set, which is composed by Android application permissions tab in the configuration file and application source code Android sensitive API, then we use information gain and chi-square test combining algorithm for data preprocessing. The second point is the improvement of the characteristic properties of weighting coefficients, as such improve the detection performance of Naive Bayes classification algorithm of Android malware. Since the contribution of each feature property classification is different, so the weight of each feature attribute weighting factor is taken into account, including solving the weight coefficient is calculated using information to gain. At the same time we design Android malware detection framework based on improved Naive Bayes algorithm, experiments are analyzed by illustrating the detection framework. Experiments show that detection Android malware based on Na?ve Bayesian classification algorithm can effectively improve the detection rate of malware and reduce the false alarm rate.