Investigation of Malware & Threat Analysis on APKs Using SVM & ANN Algorithm. -A New Approach

With 85% market share, Android is currently the one of the most popular smart-mobile device platform around the world. The popularity of Android encourages cyber criminals to produce harmful apps that might jeopardize the security and confidentiality of the mobile devices. Every year almost above 55% of apps are being rejected by google play store. Google has banned more than 22 harmful Adware applications from the google play store in the previous year, which added up to around at least 7.5 million downloads. In the year 2017, they had to remove around 700,000 applications that broke the standards sets by google play, and almost more than a million were identified as bad developers. There has always been struggle against malware it is never-ending process, and at the end it is always the user who has their significant number of personal and private data stolen by attackers, jeopardizing their privacy and leaving them vulnerable to additional threats. To tackle this increasing rate of malware production, a malware detection system that is both efficient and flexible is required. There are three primary ways in detection of android malware, they are static analysis, dynamic analysis, and hybrid. In this paper, we will look into malware detection techniques based on SVM and ANN for android applications. Here, we first compute the similarity scores between malware and benign applications in terms of suspicious API calls and use these similarity scores as features in the feature vector, then we train the SVM classifier with different risky permissions combinations as additional features for training. This model will demonstrate that the suggested technique can properly detect whether a particular program is malicious or not.