Android Malware Detection Based on Static Behavior Feature Analysis

As an open source operating system, Android has groups of users and developers, which result in a great deal of Android application in app markets. But at the same time, app’s benign and malicious behavior cannot be screened, the number of new malware sample for Android platform and infected mobile phone are still soaring, Android malware detection is faced with a tough challenge. Thus this paper proposes an approach that analyze Android malware by extracting app’s static behavior, and information are captured include the use of permission, android’s components and API calls. Then we take the advantage of PCA to extract app’s principle behavior features. Finally, classifiers are trained by Android app dataset to validate the performance. Experiments show that the proposed method is capable of 97.4% true positive rate and 99.8% area under ROC, our lightweight classifiers can detect Android’s unknown malware effectively and accurately.