No abstract

Mobile phones has become very integral part in our day to day life. In the digitalized world most of our day to day activities rely on mobile phone like banking activities, wallet payments, credentials, social accounts etc. Our system works in such a way that if there is an advantage to a technology there also exists a disadvantage. Every users have all their private and sensitive data in their mobile phones and download random applications from different platforms like play store, App store etc. There is a huge possibility that the applications downloaded are malicious applications. The existing system provides a solution for detection of such applications with the help of antivirus which has pre-built signatures that can be used to obtain an already existing malware which can be modified and manipulated by the hacker if they tend to do so. In this project, our purpose is to identify the malicious applications using Machine learning. By combining both static analysis and dynamic analysis we can use a Hybrid approach for analysing and detecting malware threats in android applications using Recurrent Neural Network (RNN). The main aim of this project will be to ensure that the application installed is benign, if it is not, it should block such applications and notify the user.

Android applications are available in the google cloud apps market. Starting from the normal functionalities like calling, messaging, and camera to advanced functionalities like online banking, online shopping, there is no limit to how we can make use of mobile phones. Just like there is no limit to functionalities of the mobile phone, there is no limit to the amount of information available on mobile phones. When you say information, it is confidential information including personal information, username, and password and card details. There are already many cases reported about information leakage by compromising a mobile phone. The important point to be noticed here is that the medium through which mobile phones are providing these functionalities, and that medium is called an application. There are millions of applications in the Play Store and App Store which come with different functionalities. The only way to stop this problem is by stopping the user from downloading malicious applications from the Play Store. But the main challenge in this solution leads to a question, which is "how the user will distinguish between a malicious application and benign application". There are millions of applications in play Store and considering play Store is a trusted media, it is very unlikely to raise any suspicion over an application to a normal user. We propose a novel approach to perform static and dynamic analysis of malicious payload and compare it with the genuine application.