A Cross Language Code Security Audit Framework Based on Normalized Representation

Abstract

With the rapid development of information technology, audit objects and audit itself are more and more inseparable from software. As an important means of software security audit, code security audit will become an important aspect of future audit that cannot be ignored. However, the existing code security audit is mainly based on source code, which is difficult to meet the audit needs of more and more programming languages and binary commercial software. Based on the idea of normalized transformation, this paper constructs a cross language code security audit framework (CLCSA). CLCSA first uses compile/decompile technology to convert different high-level programming languages and binary codes into normalized representation, and then uses machine learning technology to build a cross language code security audit model based on normalized representation to evaluate code security and find out possible code security vulnerabilities. Finally, for the discovered vulnerabilities, the heuristic search strategy will be used to find the best repair scheme from the existing normalized representation sample library for automatic repair, which can improve the effectiveness of code security audit. CLCSA realizes the normalized code security audit of different types and levels of code, which provides a strong support for improving the breadth and depth of code security audit.