Another free app: Does it have the right intentions?

摘要: Security and privacy holds a great importance in mobile devices due to the escalated use of smart phone applications (app). This has made user even more vulnerable malicious attacks than ever before. We aim address this problem by proposing novel framework identify potential Android malware apps extracting intention their permission requests. First, we constructed dataset consisting 1,730 benign along with 273 samples. Then, both datasets were subjected source code extraction. From there on, followed two phase approach In 1, machine learning model group into different clusters based on operations known as task-intention. Once trained model, it was used task-intention an app. Further, phase, only construct task-intentions none signatures involved. Therefore, our does not models apps. for each group, extracted permission-requests probability mass functions (PMF). named shape PMF Intention-Shape or I-Shape. 2, permission-requests, I-Shapes compared unknown app its corresponding I-Shape Using approach, obtained accuracy 89% detecting The novelty work is perform identification without training any signatures, utilization such Our can be utilized safety before installed performs static analysis. pre-screening multi-layer security systems. It also highly useful screening when launching markets.