Our micro-benchmark suite DroidBench (published with FlowDroid at PLDI’14) aims at testing the precision and recall of static taint tracking tools for Android. It provides categorized, tested, and well-documented test cases for the various hard challenges in program analysis. The ground truth is provides makes it easy to check and compare the results of the various information-flow analysis tools proposed both in research and available commercially.
The suite has been used by various research groups all over the world and we have seen tools greatly improve on the precision and recall they achieve on DroidBench. With many tools now achieving very good results, it is time for new challenges.
We are thus happy to announce that DroidBench 2.0 is now available from Github. It features 120 test cases in 13 categories including aliasing, implicit data flows, Android lifecycle handling, inter-component communication, and reflective method calls. We would like to thank all the researchers world wide that have contributed to DroidBench and would like to extend this call: Feel free to propose and/or submit new test cases to extend the suite even further so that it can continue to serve as a standardized benchmark suite for research in the field of static taint tracking.
All kinds of contributions are welcome. We have started to also add test cases challenging dynamic analysis tools, for instance emulator-detection mechanisms. In the future, we also plan to add test cases that leverage native code to hide data flows.