Permission and API Calls Based Hybrid Machine Learning Approach for Detecting Malicious Software in Android System

Since last decade the usage of smart phones, PDAs, tablets have become increased due to its usability, friendliness and portability. Because of this extensive growth, mobile malware targeting the android platform. Android users store personal information like name, address, contacts, text message, financial transactions, one time password, mail IDs, net banking details, making calls without user authentication, recording voice calls etc. Mobile devices offered lot of services and application than those offered by other personal computer. In this research paper, we have done a combined malware detection tool and analyzer and removal tool for group of application by considering computational complexity, redundancy and detection ratio of mobile malware. As a detailed summary, the research in this paper focus on makes significant contributions for android malware detections by means of following steps: 1. The process of extracting source code from mobile android apk file. 2. We create a array of structures from extracted information of mobile android application in order to develop a mobile malware detection framework.2. Finally create a database by removing redundancy which will be useful for identifying malware / fault information. Support Vector machine learning algorithm and logistics regression are used for differentiating Benign from Malware app. Finally calculating accuracy in terms of precision, accuracy and recall. The experimental results show the true positive classification of malicious at 95.5% and false positive rate as 0.8%.