XML Depth-First Resolution (.//icon_url) + Internal Fetch via pycurl

.//field style lookups are tree-wide and depth-first. The first nested match, not the visually obvious top-level field, may become the trusted value. Validation often inspects one node while later fetch logic resolves another — or both use tree-wide queries and the attacker controls ordering.

Why It Works

• Generic selectors like .find(".//icon_url") return the first match in document order, which the attacker can plant beneath a benign-looking wrapper.

Vulnerable Pattern

• Python XML parsing with .find(".//icon_url"), .find(".//method"), or similar generic queries over attacker-controlled XML.
• The fetched URL or HTTP method is later passed straight into pycurl, requests, or another backend call.

Exploit Flow

1. Discover which fields are read with tree-wide selectors and whether the first match wins.
2. Build a document where a benign top-level field satisfies human review while the first depth-first match contains the real payload.
3. Target internal proxy hosts, metadata endpoints, or method/body controls to pivot beyond simple SSRF.

Variations

• Nested wrappers, duplicated field names, namespace tricks, or one malicious field paired with a validation-safe twin.

Common Blockers

• Schema validation enforcing exact node positions, or code that reserializes the XML before use.

PoC Sketch

<root>
  <meta><icon_url>http://127.0.0.1/internal</icon_url></meta>
  <icon_url>https://cdn.example/logo.png</icon_url>
</root>

Good Situations To Use It

• An app parses attacker XML with .// selectors.
• The same field feeds a server-side fetch.
• You can nest elements to control depth-first ordering.

Sources

• fcsc2026/web/bubulle_corp
• fcsc2026/web/bubulle_corp_2