Executive brief
Rucio, a scientific data management system, contains a critical security flaw in how it handles searches when using an Oracle database. An authenticated user can bypass security restrictions to view, modify, or delete any data in the system's database, including sensitive credentials and password hashes. This could lead to a total compromise of the data management platform and unauthorized access to managed storage resources.
Technical details
A SQL injection vulnerability exists in `FilterEngine.create_sqla_query` within `lib/rucio/core/did_meta_plugins/filter_engine.py`. When using the Oracle dialect, the application uses Python's `str.format` to interpolate attacker-controlled filter keys and values directly into `sqlalchemy.text()` fragments, bypassing SQLAlchemy's standard parameterization. The vulnerability is reachable via the `GET /dids/<scope>/dids/search` endpoint. An authenticated attacker can use this to perform boolean-based or time-based blind SQL injection to extract sensitive data such as session tokens and password hashes, or bypass scope-based access controls. This issue specifically affects deployments using the `json_meta` plugin with Oracle; PostgreSQL and MySQL deployments are not affected as they use proper bind parameters. Patches are available in versions 35.8.5, 38.5.5, 39.4.2, and 40.1.1.
Affected products
- rucio rucio >= 1.27.0, < 35.8.5; >= 36.0.0, < 38.5.5; >= 39.0.0, < 39.4.2; >= 40.0.0, < 40.1.1
Timeline
- 2026-05-06: advisory
- 2026-05-06: disclosed