Date of Publication : July 2026

Abstract: Android is a growing platform with applications that have spurred a rise in threat of malware attacks, making Android security an important topic for study. Static analysis techniques are not effective against modern malware that uses code obfuscation, code encryption, packing and dynamic code loading. Dynamic analysis has proven to be a good method of detecting Android malware in runtime by observing the behavior of applications and detecting malicious activities that might not be visible in static code. This paper presents a comprehensive survey of dynamic analysis approaches for Android malware detection. Different dynamic analysis methods such as sandbox analysis, system call analysis, API call analysis, taint analysis, behavioral analysis and machine learning based are discussed and compared. The study also looks at some of the major dynamic analysis frameworks and recent deep learning-based Android malware detection models that have been developed. In addition, the benefits, constraints, and problems with dynamic analysis, including anti-emulation methods, partial execution coverage, high execution time and scalability issues, are reviewed. Finally, novel research directions such as Explainable Artificial Intelligence, Federated Learning, Cloud-based analysis and Adversarially Robust Detection Models are highlighted. The survey offers valuable insights about the current progress and future possibilities in Android malware detection using dynamic analysis techniques.

Reference :