摘要： 基于传统模糊测试框架的S7协议模糊测试技术存在构造困难和代码覆盖率低的问题，对测试效率和质量产生很大影响。借助神经网络模型对数据较强的学习能力和预测能力，提出一种基于长短期记忆（LSTM）神经网络的S7协议模糊测试用例生成方法。将S7协议中的特征值字段分为可变字段和不可变字段，对可变字段进行模糊处理，对不可变字段做固定值操作，进而利用局部模糊实现对S7协议帧各字段的模糊分析，生成有效的测试用例。经过学习，模型可以提取到西门子S7协议的特征，自动产生满足协议结构的测试用例。实验对不同字段进行局部模糊，结果表明，该方法预测出的数据具备真实测试用例的特征，可生成大量对特征字段关联性较大的有效测试用例，提高代码覆盖率。

关键词: 长短期记忆神经网络, S7协议, 模糊测试, 测试用例, 字段

Abstract: The fuzzy test technology for S7 protocol based on the traditional fuzzy test framework is limited by the difficulty in construction and low code coverage,which has a great impact on the test efficiency and quality.Exploiting the strong learning and prediction ability of the neural network model,a Long Short-Term Memory(LSTM) neural network-based method of generating fuzzy test cases for S7 protocol is proposed.The feature value fields in the S7 protocol are categorized into the mutable fields and the immutable fields.The mutable fields are blurred,and the immutable fields are specified with fixed values.And then local fuzzy is used to perform fuzzy analysis on each field of the S7 protocol frame to generate effective test cases.The learned model can extract the features of the Siemens S7 protocol and automatically generate test cases conforming to the protocol structure.Local fuzzy is performed on different fields for experiments,and the results show that the generated data has the features of real test cases.The method can provide a large number of effective test cases with strong correlation with the feature fields,improving the code coverage.

Key words: Long Short-Term Memory(LSTM) neural network, S7 protocol, fuzzy test, test case, field

中图分类号: 

• TP391.1