基于双向LSTM的软件漏洞自动识别方法研究

Research on Automatic Software Vulnerability Identification Method Based on Bidirectional LSTM

传统软件漏洞自动识别方法的识别结果不够理想,为此提出基于双向LSTM的软件漏洞自动识别方法。该方法利用开源工具ANTLR4建立抽象语法树获取软件源代码特征,作为软件漏洞的识别依据;利用双向LSTM技术对特征数据进行训练分析,检测到软件运行中存在的漏洞,以此实现基于双向LSTM的软件漏洞自动识别。实验结果证明,设计方法的相似度高于传统方法。

The recognition results of traditional software vulnerability automatic identification methods are not ideal. For this reason, a software vulnerability automatic identification method based on two-way LSTM is proposed. This method uses the open source tool ANTLR4 to build an abstract syntax tree to obtain software source code features, as the basis for identifying software vulnerabilities;uses two-way LSTM technology to train and analyze feature data, and detects vulnerabilities in software operation, so as to achieve a two-way LSTM Automatic identification of software vulnerabilities. The experimental results prove that the similarity of the design method is higher than that of the traditional method.