Towards More Security in Data Exchange: Defining Unparsers with Context-Sensitive Encoders for Context-Free Grammars
This addresses security issues for developers and systems handling context-free language documents, though it is incremental as it builds on existing grammar-based approaches.
The paper tackles the problem of injection vulnerabilities like XSS and SQL injection in distributed systems by proposing a method to derive context-sensitive encoders from context-free grammars, ensuring correct unparsing of malicious input data for all context-free languages, resulting in documents where input data does not influence structure or semantics.
To exchange complex data structures in distributed systems, documents written in context-free languages are exchanged among communicating parties. Unparsing these documents correctly is as important as parsing them correctly because errors during unparsing result in injection vulnerabilities such as cross-site scripting (XSS) and SQL injection. Injection attacks are not limited to the web world. Every program that uses input to produce documents in a context-free language may be vulnerable to this class of attack. Even for widely used languages such as HTML and JavaScript, there are few approaches that prevent injection attacks by context-sensitive encoding, and those approaches are tied to the language. Therefore, the aim of this paper is to derive context-sensitive encoder from context-free grammars to provide correct unparsing of maliciously crafted input data for all context-free languages. The presented solution integrates encoder definition into context-free grammars and provides a generator for context-sensitive encoders and decoders that are used during (un)parsing. This unparsing process results in documents where the input data does neither influence the structure of the document nor change their intended semantics. By defining encoding during language definition, developers who use the language are provided with a clean interface for writing and reading documents written in that language, without the need to care about security-relevant encoding.