Abstract: Currently, in the smart-phone market, third-party stores are contributing a major share in developing Android apps. This paper represents a technique designed to calculate risk-factor for applications installed on smart-phones via third-party stores. The system represents an alternative to the current detection techniques by alerting users with risk alarm signals about malicious behaviour of application for the safety of mobile devices. The implementation methodology initially collects mobile applications from third-party stores and extracts feature set from many parameters of running mobile applications in an emulated environment. After that, for each feature-set, algorithms are proposed to calculate the risk factor for all the parameters which reveals the risk-level and its malicious impacts on mobile phones and leakage of privacy. Finally, the system is tested on four learning algorithms with WEKA API to find the best classifier for classifications of risk-levels. Among them, Logistic Regression shows 96% accuracy, RBF(Radial Basis Function) show 99% accuracy with some false positives, SMO (Sequential Minimal Optimization) shows 96.6% accuracy and Naïve Baye’s produces 99.8% accurate results with very low false positives.  Therefore, it is concluded that Naïve Baye’s classifier can be integrated with the devised technique in future to detect risk levels of third-party applications.

Keywords: Smart Devices; android applications; malware; network traffic; permissions; risk-factor; logistic regression; Naïve Baye’s; radial basis function

PDF | DOI: 10.17148/IJARCCE.2019.8919

Open chat
Chat with IJARCCE