EspyDroid+: Precise Reflection Analysis of Android Apps

Abstract Malicious smartphone apps use reflection APIs to exfiltrate user data and steal personal information. These malware use reflection along with parameter obfuscation and encryption to evade detection by static analysis. Dynamic analysis is a possible approach to detect such run-time malicious behavior. However, dynamic analysis of a software, usually, results in the exploration of a large, potentially exponential, number of program branches. Many of these program paths are not useful to analyze the reflection APIs, and significantly affect the efficiency of the dynamic analysis. In this paper, we propose a hybrid analysis approach named EspyDroid+ 1 that overcomes the drawbacks of static analysis in analyzing the obfuscated and run-time dependent parameters of reflection APIs. EspyDroid+ incorporates Reflection Guided Static Slicing (RGSS), an efficient approach to deal with exploration of large number of program paths by pruning irrelevant program paths and ensures that the resultant paths get executed during the subsequent dynamic analysis. We observed that EspyDroid+ successfully removed 59.91% of the total paths on a test dataset consisting of 660 apps without any loss of semantics. We conclude that EspyDroid+ is effective, fast, and scalable in uncovering reflection API induced privacy leaks.