Add Support for HTTP Headers in URL Fetch Requests with Secure Storage for Landing Requests

[davelopez]

Currently, we can only fetch data from public URLs without any authentication or custom headers.

This PR introduces support for HTTP headers in URL fetch requests for landing requests. Headers are controlled through pattern-based configuration, and sensitive headers are automatically encrypted using Galaxy's vault system before storing in the database.

🚀 Features

1. Pattern-Based URL Header Configuration

• URL Matching: Headers controlled per URL pattern using glob syntax (*, ?, **)
• Multiple Patterns: One URL can match multiple patterns - union of all allowed headers is permitted
• Configuration File: config/url_headers_conf.yml
• Fail-Fast Security: Headers rejected if used without proper configuration

2. Automatic Sensitive Header Encryption

• Per-Pattern Configuration: Each pattern explicitly declares which headers are sensitive
• Vault Integration: Sensitive headers encrypted using Galaxy's vault system
• Transparent Operation: Non-sensitive headers remain in plain text for performance
• Secure-by-Default: Header is sensitive if ANY matching pattern marks it sensitive

3. Secure Storage Architecture

• Database Protection: Sensitive header values are never stored in plain text
• Vault References: Encrypted headers replaced with vault placeholders (e.g., __VAULT_HEADER_AUTHORIZATION__)
• Automatic Decryption: Headers are automatically decrypted when landing requests are retrieved
• Key Management: Hierarchical vault keys: headers/{landing_uuid}/{header_name}

🔧 Configuration

Example Configuration File

patterns:
    - url_pattern: "https://github.com/**"
      headers:
          - name: "Authorization"
            sensitive: true
          - name: "Accept"
            sensitive: false
    - url_pattern: "https://api.example.com/v1/**"
      headers:
          - name: "X-API-Key"
            sensitive: true

How Pattern Matching Works

• All-Matches Logic: If a URL matches multiple patterns, the union of all allowed headers is permitted
• Glob Syntax: Standard glob patterns (* = any chars, ? = single char, ** = recursive)
• Order Independent: Pattern order doesn't matter - all matches contribute to allowed headers

🔧 How It Works

API Usage Examples

Creating a Data Landing Request with Headers

POST /api/data_landings
Content-Type: application/json

{
  "request_state": {
    "targets": [{
      "destination": {"type": "hdas"},
      "items": [{
        "src": "url",
        "url": "https://api.example.com/data.json",
        "ext": "json",
        "headers": {
          "Authorization": "Bearer secret-token-123",
          "X-API-Key": "api-key-456",
          "User-Agent": "Galaxy",
          "Content-Type": "application/json"
        }
      }]
    }]
  },
  "public": true
}

Creating a Workflow Landing Request with Headers

POST /api/workflow_landings
Content-Type: application/json

{
  "workflow_id": "workflow_123",
  "workflow_target_type": "stored_workflow",
  "request_state": {
    "input_dataset": {
      "src": "url",
      "url": "https://secure-data.example.com/dataset.csv",
      "ext": "csv",
      "headers": {
        "Authorization": "Bearer workflow-token-789",
        "X-Custom-Header": "custom-value"
      }
    }
  },
  "public": true
}

Under the Hood: Encryption Process

1. Configuration Check: System validates headers against configured URL patterns
2. Pattern Matching: All matching patterns identified for the URL
3. Header Validation: Only headers allowed by at least one matching pattern are accepted
4. Sensitivity Detection: The Header is sensitive if ANY matching pattern marks it sensitive
5. Vault Storage: Sensitive headers encrypted and stored in vault:

   landing_request/headers/{landing_uuid}/authorization
   landing_request/headers/{landing_uuid}/x_api_key
   

6. Reference Replacement: Sensitive values replaced with vault references:

   {
       "headers": {
           "Authorization": "__VAULT_HEADER_AUTHORIZATION__",
           "X-API-Key": "__VAULT_HEADER_X_API_KEY__"
       }
   }

7. Transparent Decryption: When a landing request is retrieved, vault references are automatically replaced with actual values

🔒 Security Features

Pattern-Based Access Control

• Explicit Allowlist: Only headers explicitly configured for matching URL patterns are allowed
• Fail-Fast: Requests with unauthorized headers are rejected immediately
• Secure-by-Default: If any matching pattern marks a header as sensitive, it's treated as sensitive

Vault Configuration Required

This feature requires a configured Galaxy vault. See the vault documentation for setup instructions.

Fallback Behavior

• No Configuration: Missing config file returns null configuration (no headers allowed)
• No Vault: Feature requires vault - fails fast if vault not configured and sensitive headers are used

✅ Testing

• Included unit and integration tests covering:
  • Pattern matching logic
  • Header validation and sensitivity detection
  • Vault encryption/decryption process
  • API request handling with headers

🎯 Use Cases

This enhancement enables several important use cases:

1. Private API Access: Fetch data from APIs requiring authentication tokens
2. Rate Limiting: Include API keys for higher rate limits
3. Custom Protocols: Support for proprietary authentication schemes
4. Workflow Integration: Secure data fetching within workflow executions

How to test the changes?

• I've included appropriate automated tests.
• This is a refactoring of components with existing test coverage.
• Instructions for manual testing are as follows:
  1. [add testing steps and prerequisites here if you didn't write automated tests covering all your changes]

License

• I agree to license these and all my past contributions to the core galaxy codebase under the MIT license.

[jmchilton]

I would rather this fail outright than risk storing things that should be encrypted in an unencrypted fashion - especially given the rest of the app will assume the encryption has already occurred. Does this make testing harder or something?

[davelopez]

The new version should cover this. Thank you!!

[jmchilton]

This approach has made the admin configuration trivial and deployment much easier as a result. The existing upload process as is is already... sort of exploitable... I mean we don't do a great job at rate limiting Galaxy (maybe this has improved?) and we let most URIs be accessed for users on behalf of the Galaxy server - it is scary from a security perspective. Allowing users to set arbitrary headers including (especially?) user-agent makes it even a richer target for hacking I would suspect. If we shipped an allow-list of headers and URI patterns that allow that header and whether the header should be secured then I would be much more comfortable from a security perspective. It would be much harder to configure then but we would be sure exactly what the exploit surface is.

Additionally, I trust the list of headers is relatively complete and well thought through but again I would be more comfortable if we had an explicit allow list because again we would understand the exploit surface exactly.

I'm not a -1 on any of this though - I'm just expressing my concerns and telling you how I would have had it work - which people may think would be too much config. Still though one can imagine blends of the approaches - maybe it is off by default but there is a configuration that allows any requests like this or restricted set of requests and admins can decide on their level of comfort.

Even if people believe I'm being too cautious or appropriately cautious but the admin/deployment burden of addressing it would be too steep - I would still strongly encourage we don't allow the user agent to be overridden - if an API wants Galaxy to access it they shouldn't require a non-Galaxy user agent.

What we pick to allow through makes me anxious - but after that - the actual implementation of allowing those headers and securing seems really well thought through well. It seems to fit with our existing APIs beautifully - that part of the implementation seems perfect to me.

[davelopez]

Thank you @jmchilton! As always, great constructive feedback! I will add some configuration to make this functionality more explicitly controlled 👍

[davelopez]

I've added the config option to specify URL patterns and sets of allowed headers for each pattern. This should be more explicit while maintaining flexibility for admins to allow general safe headers. I've also updated the PR description with the updates.
Thanks again for the feedback! Let me know if there is something else worth improving 🙏