To content
Fakultät für Informatik

New Benchmark for Android Taint Analysis

An overview diagram of the TaintBench framework © Luo, Pauck, Piskachev, Benz, Pashchenko, Mory, Bodden, Hermann, Massacci
We provide a real-world benchmark to evaluate Android taint analysis. The paper was accepted for the Empirical Software Engineering Journal.

While some benchmarks based on small self-written example code for Android taint analysis such as DroidBench existed before, there was no benchmark composed out of real-world code. This impeded the reliability of the evaluations in this field as it made them restricted and hard to compare. In this new paper we provided a novel benchmark based on Android malware and a framework to created and evaluate such benchmarks.  The documented taint flow baseline was established through a rigorous process. We evaluate our benchmark and methodology through several experiment involving taint analysis tools and user studies. 

The Empirical Software Engineering journal is a top publication venue in the field of software systems. With an impact factor of 2.522 and an h5-index of 52 Google Scholar ranks it #6 in the field. We are happy to provide our paper open access under the German DEAL agreement.

Evaluation of Amadroid and FlowDroid with the DroidBench and TaintBench benchmarks © Luo, Pauck, Piskachev, Benz, Pashchenko, Mory, Bodden, Hermann, Massacci
Evaluation of Amandroid and FlowDroid with the DroidBench and TaintBench benchmark suites

TaintBench: Automatic Real-World Malware Benchmarking of Android Taint Analyses
Linghui Luo, Felix Pauck, Goran Piskachev, Manuel Benz, Ivan Pashchenko, Martin Mory, Eric Bodden, Ben Hermann, and Fabio Massacci
In Empirical Software Engineering, Springer Heidelberg
DOI: https://doi.org/10.1007/s10664-021-10013-5
Artifact: https://taintbench.github.io