Android Malware Identification Based on Traffic Analysis

As numerous new techniques for Android malware attacks have growingly emerged and evolved, Android malware identification is extremely crucial to prevent mobile applications from being hacked. Machine learning techniques have shown extraordinary capabilities in various fields. A common problem with existing research of malware traffic identification based on machine learning approaches is the need to design a set of features that accurately reflect network traffic characteristics. Obtaining a high accuracy for identifying Android malware traffic is also a challenging problem. This paper analyses the Android malware traffic and extract 15 features which is a combination of time-related network flow feature and packets feature. We then use three supervised machine learning methods to identify Android malware traffic. Experimental results show that the feature set we proposed can accurately characterize the traffic and all three classifiers achieve high accuracy.