提交 #791920: Langflow <= 1.8.3 Information Disclosure

# Technical Details
An Information Disclosure vulnerability exists in Langflow. The application fails to properly redact sensitive credentials when exporting flow JSON files due to flawed logic in the `remove_api_keys` utility.

The application only strips values if the field is marked as `password=True` AND its name strictly contains "api" AND ("key" or "token"). If a sensitive field is just named `password`, `secret_value`, or `db_password`, the redaction logic is completely bypassed, leaking the plain-text secret.

# Vulnerable Code
File: `src/backend/base/langflow/api/utils/core.py`
Method: `remove_api_keys()` and helper `has_api_terms()`
Why: Custom or generic sensitive fields (like database passwords or unrelated secrets) that do not match the restrictive `has_api_terms` naming heuristic are ignored by the redactor, even if they are explicitly flagged with `password=True` in the schema.

# Reproduction
1. An attacker (or regular user) creates a flow containing a sensitive field (e.g. named "password") with `password=True`.
2. The user stores a highly sensitive string such as "SUPER_SECRET_VALUE" in this field.
3. The user downloads the flow using the API (`POST /api/v1/flows/download/`).
4. The downloaded JSON file reveals the raw plain-text secret "SUPER_SECRET_VALUE", as the `remove_api_keys` function wrongly decided to leave it intact.

# Impact
- Exposure of Sensitive Information: Database credentials, third-party service secrets, and other confidential data are leaked in exported JSON files.
- Unauthorized actors gaining access to these flow exports can extract plaintext credentials and pivot to attack integrated external infrastructure.

• 来源: ⚠️ https://gist.github.com/chenhouser2025/b93261c6e651f14800a4f2e4365f357b
• 用户: Eric-d (UID 96861)
• 提交: 2026-03-28 14時36分 (23 日前)
• 管理: 2026-04-19 15時47分 (22 days later)
• 状态: 已接受
• VulDB条目: 358232 [langflow-ai langflow 直到 1.8.3 Flow Using API core.py remove_api_keys/has_api_terms 权限提升]
• 积分: 20