SMELLDroid: A Dataset for Code Smells in Android Apps
Year: 2026
Type: conference-paper
Abstract: Code smells are recurring design and implementation issues that degrade software quality and maintainability. While several studies have examined code smells in Android applications, there is a lack of publicly available datasets that systematically capture and quantify them. This paper introduces SMELLDroid, a large-scale dataset designed to support empirical studies on code smells within the Android ecosystem. The dataset comprises code smells extracted from 38,704 Android applications, including 29,201 malware apps and 9,503 benign apps. SMELLDroid encompasses seven code smell types—four object-oriented (BLOB, CC, LM, SAK) and three Android-specific (HAS, HBR, HSS)—with quantitative indicators representing their occurrence across applications. This dataset enables researchers to pursue multiple directions, such as quantifying smell prevalence, analyzing relationships among smell categories, and comparing code quality characteristics between benign and malware apps.
Authors: Joyce Champie, Karim Elish, Mahmoud Elish
Institution: Florida Polytechnic University
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7171941663