This repository is the accompaniment to the paper: On Reducing the Network Usage of Android Applications with Genetic Improvement, accepted to the 13th International Workshop on Genetic Improvement @ ICSE 2024 (https://geneticimprovementofsoftware.com/events/icse2024).

In this repository is the code to instrument http requests made by Android applications, a modified version of the GIDroid framework (https://github.com/SOLAR-group/GIDroid), and the results of our experiments.

Requires Java 11 and maven.

To instrument applications, simply build a jar using the command:

clean compile assembly:single

Next call the jar with arguments a (path to apk to instrument) and p (the name of the package in which the classes you wish to instrument are.

Once instrumented, simply install the apk onto a device or emulator, and run a random input generation tool, such as Monkey (https://developer.android.com/studio/test/other-testing-tools/monkey) on your code. All http request outputs will be logged at the debug level.

Full instructions for running GIDroid are available in the readme in the GIDroid directory.

Results from our experiments are available in the results directory, they are divided by which search algorithm was used and then the application that GIDroid ran on.