一句话：提出 GDSynth，将元模型变异与领域特定图模型合成结合，为图数据库自动生成多样测试数据并发现真实缺陷。
图数据库管理系统（GDBMS）被广泛用于包括安全关键在内的多个领域，而测试此类系统需要大量且多样的测试数据，即查询与图内容的组合。然而现有 GDBMS 测试工具对合成有意义且多样的测试图支持十分有限，制约了测试的有效性。本文提出 GDSynth 测试框架，创新性地将（1）元模型变异与（2）领域特定图模型的自动合成相结合，把合成出的图模型作为测试数据在已有测试查询上执行。作者在 Neo4J 图数据库上开展自动化蜕变测试（metamorphic testing）以展示框架的实用价值，并成功揭示了一个此前已被报告的真实缺陷。演示视频与开源代码均已公开。
Graph database management systems (GDBMSs) are used in various domains, including safety-critical applications. A challenging aspect of testing such systems is that it needs a large and diverse amount of test data consisting of queries and graph contents. However, existing GDBMS testing tools provide very limited support for synthesizing a meaningful and diverse set of test graphs. This paper presents GDSynth , our novel testing framework that aims to reveal bugs in graph database management systems by innovatively combining (1) metamodel mutations with (2) automated synthesis of domain-specific graph models as test data executed on existing test queries. The usefulness of our framework is showcased by performing automated metamorphic testing of the Neo4J graph database and revealing a previously reported real bug. Our demo video is available at https://youtu.be/dIPlf0G3-d0 and our open-source code is available at https://github.com/FozAhm/GDSynth.
@inproceedings{paperbot11754,
title = {GDSynth: A Graph Database Testing Framework via Effective Graph Synthesis},
author = {Fozail Ahmad and Kristóf Marussy and Oszkár Semeráth and Dániel Varró and Lili Wei},
booktitle = {ICSE},
year = {2026},
doi = {10.1145/3774748.3787605}
}