The input is a list of failure logs obtained from the fuzz testing on an Android component, and the output is a set of index sets, of which each index set tells occurrences of a unique failure.
Due to the nature of fuzz testing, the existing Intent fuzzing tools must produce many duplicate failures as well.
We evaluated our tool in terms of the execution time for Intent fuzz testing, which have been rarely discussed by the existing researches.
10 shows the execution time for Intent fuzz testing, which is the time for running Intent test cases plus the time for grouping.
The time for Intent fuzz testing on a single Android app is found to be proportional to the number of Intent test cases.
Second, our Intent fuzz testing tool itself is robust to run fully automatically, which is important in practice particularly when it is applied to many Android apps by batch processing.
In summary, this testing time analysis suggests that some consideration should be taken for efficiency of Intent fuzz testing.
Despite the nature of randomness in Intent fuzz testing, our tool discovered an interesting failure that the other researches had not reported before.