A Formal Model of Checked C

• Published in: 2022 IEEE 35th Computer Security Foundations Symposium (CSF) pp. 49 - 63
• Main Authors: Li, Liyi, Liu, Yiyun, Postol, Deena, Lampropoulos, Leonidas, Van Horn, David, Hicks, Michael
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2022
• Subjects: Analytical models; Annotations; C Semantics; Checked C; Codes; compiler verification; Computational modeling; Generators; Memory Safety; Safety; Semantics; Spatial Safety; static analysis; type soundness
• DOI: 10.1109/CSF54842.2022.9919657

We present a formal model of Checked C, a dialect of C that aims to enforce spatial memory safety. Our model pays particular attention to the semantics of dynamically sized, potentially null-terminated arrays. We formalize this model in Coq, and prove that any spatial memory safety errors can be blamed on portions of the program labeled unchecked; this is a Checked C feature that supports incremental porting and backward compatibility. While our model's operational semantics uses annotated ("fat") pointers to enforce spatial safety, we show that such annotations can be safely erased. Using PLT Redex we formalize an executable version of our model and a compilation procedure to an untyped C-like language, as well as use randomized testing to validate that generated code faithfully simulates the original. Finally, we develop a custom random generator for well-typed and almost-well-typed terms in our Redex model, and use it to search for inconsistencies between our model and the Clang Checked C implementation. We find these steps to be a useful way to co-develop a language (Checked C is still in development) and a core model of it.