On Precise Modeling of Regular Replacement

This paper studies the precise modeling of various semantics of regular substitution, such as the declarative, finite, greedy, and reluctant replacement, using finite state transducers (FST) as filters. By projecting an FST of regular replacement to its input/output tapes, we are able to solve atomic string constraints, which can be applied to both the forward and backward image computation in model checking and symbolic execution. We report several interesting discoveries, e.g., certain fragments of the general problem can be handled using less expressive deterministic finite state transducers. A compact representation of FST and the customized algorithms for FST operations are developed. It is shown that the precise modeling of regular replacement can help discover delicate security vulnerabilities of software systems with very few false