An(other) Exercise in Measuring the Strength of Source Code Obfuscation

摘要

We experimentally compare the strength of different source code obfuscation techniques by measuring the performance of human analysts. We describe an experimental setup by which it is possible to compare different obfuscation techniques with each other. As techniques, we considered name overloading and opaque predicates, as well as the combination of both. While the results are interesting and show that increased levels of obfuscation decrease the performance of humans, only one result (the use of name overloading) was statistically significant.