Internal machine learning API endpoint for CWE classification is vulnerable to path traversal
该报告还没有中文解读。
原文摘录
摘自原文（2395 字符中的一小部分）
HackerOne has an internal machine learning API that exposes inference endpoints for numerous machine learning / artificial intelligence solutions. In one of the endpoints, `/predict/report_weakness_id`, which is used to classify report input, a path traversal vulnerability exists that could lead to …
原文中的代码
curl -X POST http://localhost:8082/predict/report_weakness_id -H 'content-type: application/json' -d'{"version":"v1", "trained_at": "2023-01-01T00:00:00Z/../../..", "input": [{"title": "test xss", "num_of_top_predictions": 3}]}'默认不展示报告全文。以上为简短摘录，完整复现过程请阅读原文。版权归研究员与 HackerOne 所有。
时间线
- 报告提交
- 公开披露
阅读 HackerOne 原文 ↗作者 @jobert · 版权归作者与 HackerOne 所有