veraPDF Validation XXE via XFA

Summary

Description An XML External Entity Injection (CWE-611) vulnerability in veraPDF allows a remote attacker to read arbitrary files on the server file system and perform Server-Side Request Forgery by submitting a crafted PDF containing a malicious XFA stream. This affects all current versions of veraPDF-validation.

Details

The vulnerability resides in veraPDF-validation validation-model/src/main/java/org/verapdf/gf/model/impl/pd/GFPDAcroForm.java within the getdynamicRender() method. This method retrieves the /XFA entry from the PDF's /AcroForm dictionary, decodes the embedded XML stream, and parses it to extract the <dynamicRender> element value.

The vulnerability stems from the use of a default-configured DocumentBuilderFactory to parse fully attacker-controlled XML:

• The factory is created via DocumentBuilderFactory.newInstance() with no security features enabled. disallow-doctype-decl, external-general-entities, external-parameter-entities, and FEATURE_SECURE_PROCESSING are all left at their insecure defaults.
• The input passed to builder.parse() is the decoded /XFA stream taken directly from the untrusted PDF.
• The text content of the <dynamicRender> node is returned to the validation model. Note that the shipped PDF/UA-1 rule (dynamicRender != 'required') consumes this value but does not echo it into the report output, so reliable exfiltration requires the out-of-band parameter-entity technique described under Impact rather than in-band reflection.

Impact

This impacts all current releases of the veraPDF validation-model module.

Successful exploitation requires only that the target validate an attacker-supplied PDF against the PDF/UA-1 profile (or via flavour auto-detection on a PDF that declares PDF/UA-1 conformance), since getdynamicRender() is invoked by the dynamicRender != 'required' rule in the bundled PDF/UA-1 profile. No additional configuration or operator action is required.

Proposed Patch

Harden the DocumentBuilderFactory in validation-model/src/main/java/org/verapdf/gf/model/impl/pd/GFPDAcroForm.java per the OWASP XXE Prevention Cheat Sheet to disallow DOCTYPE outright.