In this paper we introduce a semantic-based approach for code obfuscation. The aim of code obfuscation is to prevent malicious users to disclose properties of the original source program. This goal can be precisely modeled by abstract interpretation, where the hiding of properties corresponds to abstract the semantics. We derive a general theory based on abstract interpretation, where the potency of code obfuscation can be measured by comparing hidden properties in the lattice of abstract interpretations. Semantic-based code obfuscation is applied to show that well known program transformation methods, such as constant propagation, can be seen as code obfuscation.