From source code to test cases: A comprehensive benchmark for resource leak detection in Android apps

Area of Science:

• Software Engineering
• Mobile Application Security
• Computer Systems

Background:

• Android applications utilize shared resources (sensors, camera, GPS) governed by usage policies.
• Failure to adhere to these policies can lead to resource leaks, where resources are acquired but not released, causing functional issues or battery drain.
• Existing research on resource leak detection lacks standardized benchmarks, hindering comparative analysis and increasing researcher workload.

Purpose of the Study:

• To establish a public benchmark, named AppLeak, for Android applications suffering from resource leaks.
• To facilitate the evaluation and comparison of diverse resource leak detection and fixing techniques.
• To reduce the burden on researchers by providing a readily available dataset.

Main Methods:

• Development of a public benchmark comprising faulty Android applications.
• Inclusion of fixed versions of the apps, where available.
• Creation of test cases for automated reproduction of resource leaks.
• Compilation of supplementary information to assist researchers.

Main Results:

• The AppLeak benchmark is publicly accessible on GitLab.
• It contains a collection of 40 distinct resource leak faults.
• The benchmark supports both static and dynamic analysis techniques for evaluation.

Conclusions:

• AppLeak provides a standardized dataset for the Android resource leak research community.
• The benchmark enables more reliable and efficient comparison of detection and fixing methodologies.
• It aims to advance the field by simplifying the verification process for new techniques.