Parser defaults can become attack paths.
XML remains common in integrations, document formats, configuration, and legacy APIs. Its flexibility is useful, but unsafe parser defaults can expose applications to serious vulnerabilities.
Common XML vulnerabilities in the wild
The most familiar class is XML External Entity injection, or XXE. When an application accepts untrusted XML and permits external entity resolution, an attacker may be able to read local files, reach internal services, or trigger denial-of-service conditions. Blind XXE can still be exploitable when the response does not contain the result, because the parser may make an observable outbound request.
Other risky features include entity expansion, XInclude, XSLT, and remote schema resolution. These behaviors can create server-side request forgery, resource exhaustion, or unexpected access to local and remote content.
What is an XML parser?
An XML parser turns markup into a structure the application can work with. Security depends on how that parser is configured, especially whether it processes document type definitions, external entities, or remote resources. The secure setting is language and library specific, so teams should verify the behavior of the exact parser and version they deploy.
Find every indirect XML input
Review every place the application accepts XML, including SOAP and SAML messages, file uploads, office documents, SVG images, RSS feeds, and backend integrations. Content-type validation alone is not enough when compressed or container formats can include XML internally.
Test behavior, not only configuration
Code review should identify parser construction and feature flags, while controlled runtime testing should confirm that external resolution is disabled. Review error behavior, outbound network activity, payload-size limits, recursion limits, and timeouts. A configuration that appears secure can be bypassed if a second parser or transformation step processes the same input later.
Secure implementation guidance
Use hardened parser settings, disable DTD and external entity processing unless there is a strict business need, prevent remote resource resolution, validate expected schemas, limit payload size, and keep parsing libraries updated. Add regression tests that fail when dangerous features are re-enabled.