Ironcladc ac库增强类型安全子集.pdfVIP

Ironclad C++ A Library-Augmented Type-Safe Subset of C++ Christian DeLozier Richard Eisenberg Santosh Nagarakatte† Peter-Michael Osera Milo M. K. Martin Steve Zdancewic Compu nd Information Science Department, University of Pennsylvania †Computer Science Department, Rutgers University {delozier, eir, posera, milom, stevez}@ santosh.nagarakatte@ 1. Introduction The C++ programming language remains widely used, de- C and C++ are widely used programming languages for im- spite inheriting many unsafe features from C—features that plementing web browsers, native applications, com- often lead to failures of type or memory safety that manifest pilers, databases, and other infrastructure software [33]. C as buffer overflows, use-after- vulnerabilities, or abstrac - and C++ provide efficiency and low-level control, but these tion violations. Malicious attackers can exploit such viola- advantages come at the well-known cost of lack of memory tions to compromise application and system security. and type safety. This unsafety allows programming errors This paper introduces Ironclad C++, an approach to such as buffer overflows (accessing location beyond the ob- bringing the benefits of type and memory safety to C++. ject or array bounds), use-after- errors (accessing mem- Ironclad C++ is, in essence, a library-augmented, type-safe