Characterizing a property-driven obfuscation strategy

Abstract

In recent years, code obfuscation has attracted both researchers and software developers as a useful technique for protecting secret properties of proprietary programs. The idea of code obfuscation is to modify a program, while preserving its functionality, in order to make it more difficult to analyze. Thus, the aim of code obfuscation is to conceal certain properties to an attacker, while revealing its intended behavior. However, a general methodology for deriving an obfuscating transformation from the properties to conceal and reveal is still missing. In this work, we start to address this problem by studying the existence and the characterization of function transformers that minimally or maximally modify a program in order to reveal or conceal a certain property. Based on this general formal framework, we are able to provide a characterization of the maximal obfuscating strategy for transformations concealing a given property while revealing the desired observational behavior. To conclude, we discuss the applicability of the proposed characterization by showing how some common obfuscation techniques can be interpreted in this framework. Moreover, we show how this approach allows us to deeply understand what are the behavioral properties that these transformations conceal, and therefore protect, and which are the ones that they reveal, and therefore disclose.