一句话:提出文档引导的模糊测试方法 DocTer,从自由文本 API 文档中提取深度学习 API 的输入约束以生成有效输入、更深入测试。
论文关注深度学习(DL)库 API 函数的测试问题。这类函数具有 DL 特有的输入约束,而这些约束通常仅以非结构化的自由文本 API 文档形式给出,现有约束提取技术难以有效提取,导致难以生成合法输入并进行深入测试。作者提出 DocTer,一种文档引导的模糊测试(fuzzing)方法:它从 API 文档中抽取 DL 特有输入约束,并据此生成符合约束的有效输入,使 fuzzing 能更深入地触发 API 函数内部逻辑。标题表明该方法面向 DL API 函数的自动化测试。给出的摘要未报告具体实验数据、缺陷发现数量或对比结果,因此关键结果需参见论文正文;其核心贡献在于用文档指导约束抽取与测试输入生成。
Input constraints are useful for many software development tasks. For example, input constraints of a function enable the generation of valid inputs, i.e., inputs that follow these constraints, to test the function deeper. API functions of deep learning (DL) libraries have DL-specific input constraints, which are described informally in the free-form API documentation. Existing constraint-extraction techniques are ineffective for extracting DL-specific input constraints.
@inproceedings{paperbot7418,
title = {DocTer: documentation-guided fuzzing for testing deep learning API functions},
author = {Danning Xie and Yitong Li and Mijung Kim and Hung Viet Pham and Lin Tan and Xiangyu Zhang and Michael W. Godfrey},
booktitle = {ISSTA},
year = {2022},
doi = {10.1145/3533767.3534220}
}