The source code of Detecting Taint-Style Vulnerabilities in Microservice-Structured Web Applications.

@inproceedings{liu2025detecting,
  title={Detecting Taint-Style Vulnerabilities in Microservice-Structured Web Applications},
  author={Liu, Fengyu and Zhang, Yuan and Chen, Tian and Shi, Youkun and Yang, Guangliang and Lin, Zihan and Yang, Min and He, Junyao and Li, Qi},
  booktitle={2025 IEEE Symposium on Security and Privacy (SP)},
  pages={972--990},
  year={2025},
  organization={IEEE}
}

First, clone the project.

git clone ...

Then open it in IDEA and set the project SDK to JDK 17 in the project settings.

Before extracting, make sure you have Python 3.11 installed, and install the required dependencies:

cd gateway_entry_scan && pip install -r requirements.txt

Next, place the gateway YAML file of your target project into the gateway_entry_scan/input folder.

For example, gateway_entry_scan/input/youlai-mall.yaml.

Fill your OPENAI_API_KEY in gateway_entry_scan/config/llm.py.

OPENAI_API_KEY = "sk-xxxxxxxxxxx"

Run main.py to start extraction.

python main.py

Then get the entry rule at gateway_entry_scan/output/<your_project_name>.json.

For example, gateway_entry_scan/output/youlai-mall.json.

Then move the rule file to src/main/resources/entry

mv gateway_entry_scan/output/<your_project_name>.json src/main/resources/entry

Before analysis, you should prepare two folders.

The first folder is the JAR folder of your target project, containing the packaged JAR files and dependencies for each microservice.

For example

jars/
    - youlai-gateway.jar
    - youlai-auth.jar
    - oms-api.jar
    - oms-boot.jar
    - pms-api.jar
    - pms-boot.jar
    - ...

The second folder is a temporary working directory for the analysis, typically located at /tmp/<your_project_name>.

Then change the analysis config in src/main/java/Starter.

public class Starter {
    public static void main(String[] args) throws IOException {
        Timer.runAndCount(() -> {
            Config.name = "<your_project_name>"; // MUST! the same as the project name in rule file: <your_project_name>.json, e.g. in the project youlai-mall, it is youlai-mall
            Config.classpathKeywords = new String[]{"com.example."}; // package name keyword of your target project to match core classes, e.g. in youlai-mall, it can be .youlai.
            Config.jarPath = "./jars"; // path to the first folder, jar folder
            Config.targetPath = "/tmp/<your_project_name>"; // path to the second folder, temp working folder
            ...
        }, Config.name);
    }
}

Then run the analysis in IDEA by executing the Starter class as the main class, specifying a large maximum Java VM memory, such as

-Xmx40g

When the analysis is complete, you can find the taint flow results in output/microservice-taint-flows.txt.

For example, in the case of youlai-mall, the output may look like:

2 TaintFlow{<com.youlai.mall.pms.controller.app.SpuController: com.youlai.common.result.PageResult listPagedSpuForApp(com.youlai.mall.pms.model.query.SpuPageQuery)>/0 -> <com.youlai.mall.pms.service.impl.SpuServiceImpl: com.baomidou.mybatisplus.core.metadata.IPage listPagedSpuForApp(com.youlai.mall.pms.model.query.SpuPageQuery)>[8@L75] $r5 = invokeinterface $r4.listPagedSpuForApp($r0, queryParams)/1 --- VUL_ID:SQLI_Mybatis_Xml}