Type Systems for Secure Programming

安全编程的类型系统由于代码可移植性的提高，语言设计中的安全性日益受到关注。大多数语言级安全机制都是在语言设计之后才考虑的。一个广为人知的语言安全系统是Java安全体系结构，可以在JDK1.2中找到。在那里，访问控制机制是作为程序本身的代码编写的，很难确定哪些访问控制实际上是适当的。本研究的目标是为编程语言开发声明性的安全体系结构。在这个项目中，将开发一种新的静态类型系统，以保证在运行时相对于某些安全属性的安全性。该研究由两个主要部分组成，*一个新的静态类型系统，其中安全信息修饰程序类型（所谓的安全访问类型），类型系统适当地加强该信息的传播；*一个新的模块系统，它包括安全访问类型作为接口的一部分，并且程序链接将需要验证安全属性。其目的是建立一个富有表现力的、灵活的安全规程，允许满足安全检查的静态验证，从而避免运行时的安全检查。静态执行属性相对于动态执行属性的优势形成了编程语言设计和软件工程的基本支柱之一：类型本身作为程序行为的简明可读规范，并且缺少运行时错误类提供了更可靠的执行行为。在安全编程的上下文中，“更可靠”直接转化为“更安全”。这项研究的长期目标是建立一个更安全的互联网。

CCR 9988491Smith, ScottJohns Hopkins UniversityType Systems for Secure ProgrammingSecurity in language design is a rising concern due to increasedportability of code. Most language-level security mechanisms havebeen afterthoughts to language designs. One widely known languagesecurity system is the Java Security Architecture, found in the JDK1.2. There, access control mechanisms are written as code in theprogram itself, and it is difficult to determine what access controlsare actually in place. The goal of this research is to develop adeclarative security architecture for programming languages. In thisproject, a novel static type system for guaranteeing safety withrespect to certain security properties at run-time will be developed.The research consists of two main components,* a novel static type system in which security information decorates program types (so-called security access types), and the type system properly enforces propagation of this information;* a novel module system which includes security access types as part of the interface, and for which program linking will entail validating security properties.The aim is an expressive, flexible security discipline which allowsstatic verification that security checks are met, allowing run-timesecurity checks to be avoided. The advantages of static over dynamicenforcement of properties forms one of the basic pillars ofprogramming language design and software engineering: the typesthemselves serve as concise readable specifications of programbehavior, and the lack of a class of run-time errors gives morereliable execution behavior. In the context of secure programming,"more reliable" directly translates to "more secure". The long-termaim of this research is a more secure internet.