Property-Driven Code Obfuscations Reinterpreting Jones-Optimality in Abstract Interpretation

Jones-optimality determines whether a specializer improves program performances. Reinterpreting this concept in terms of the precision of an abstract interpreter means to determine whether specializing a source program is able to improve the precision of a given static analysis. In the opposite direction, a specializer failing optimality (disoptimal) would decrease the precision of the analysis when applied to the specialized code. In this paper, we exploit this reinterpretation of Jones-optimality relatively to the precision of an abstract interpreter with the aim of systematically deriving obfuscated code. In line with the idea behind Futamura’s projections, we factorize the construction of the obfuscated code by separating specialization and interpretation. An interpreter specializer is then systematically made disoptimal by means of language transduction. The result is a language agnostic code obfuscator which is able to foil any given static analyzer.