CVE-2026-53723Medium· 5.8▾ Sunlitguzzlehttp/guzzle-services' XML Request Serialization Vulnerable to XML Injection via CDATA Terminator
▾ Sunlit zone — Low / medium · no exploitation signal
impact 31.9 · likelihood 0 · exploitation 0
Need a working PoC? Pro members can cast a request and our team develops one — it lands right here.
Exploit-prediction probability, daily snapshots since Jul 7.
Disclosure to exploitation, from the record and what we observed since indexing it.
Disclosed via GHSA
Last analysed / modified upstream
0.2%
guzzlehttp/guzzle-services does not safely serialize scalar XML element values containing the CDATA terminator ]]>. The XML request serializer writes values containing <, >, or & with XMLWriter::writeCData($value). If attacker-controlled input contains ]]>, the CDATA section closes early and the remainder is interpreted as XML markup. This is an outgoing request-body integrity issue, not a response parsing issue. The attacker does not need to control the service description or schema.
Users are affected when all of the following are true:
guzzlehttp/guzzle-services to serialize outgoing requests.additionalParameters schema uses location: xml.enum, pattern, or custom filter that excludes ]]>.Applications that serialize untrusted input into location: xml request parameters can emit XML containing attacker-controlled elements outside the intended text node. Depending on the receiving service, this can alter operation semantics, smuggle privileged fields, bypass modeled parameter boundaries, or create conflicting duplicated elements. Fixed service descriptions are sufficient if they contain an XML element parameter populated from attacker-controlled input.
Users are not directly affected if they only use Guzzle Services to deserialize HTTP response bodies. Response XML parsing uses the response XML location visitor and does not invoke the vulnerable request XML serializer. Response bodies matter only in a second-order flow, such as parsing attacker-controlled response XML, storing or forwarding a parsed string value, and later using it as a location: xml request parameter.
Example fixed service description:
'DisplayName' => ['location' => 'xml', 'type' => 'string']
If an attacker-controlled display name is:
Alice]]></DisplayName><Role>admin</Role><DisplayName><![CDATA[
the vulnerable serializer can emit an injected element outside the intended DisplayName text node:
<Request><DisplayName><![CDATA[Alice]]></DisplayName><Role>admin</Role><DisplayName><![CDATA[]]></DisplayName></Request>
If the downstream service treats <Role> as meaningful, the attacker has set a field the modeled DisplayName parameter was not intended to set.
The issue is patched in 1.5.4 and later by safely splitting embedded CDATA terminators before serialization. The fix preserves the original scalar value as XML text and prevents injected nodes.
If you cannot upgrade immediately, constrain attacker-controlled XML element values with a strict enum, pattern, or custom filter that excludes ]]>, or avoid serializing untrusted data into location: xml element text until patched. Where appropriate for the service schema, XML attributes are not affected because they are written with XMLWriter attribute APIs rather than CDATA sections.
To determine whether action is needed, search service descriptions for request parameters using location: xml, including operation parameters and additionalParameters. Response-only models are not directly affected unless parsed values are reused for request serialization. For object and array parameters, review nested scalar properties because leaf element values can still be affected.
guzzlehttp/guzzle-services < 1.5.4Upgrade to a patched release:
guzzlehttp/guzzle-services 1.5.4Connected by shared product, vendor, weakness, or advisory.
CVE-2026-49214Medium· 5.3guzzlehttp/psr7 has CRLF Injection via URI Host Component
CVE-2026-48998Medium· 5.3guzzlehttp/psr7 has Host Confusion via Authority Reinterpretation
CVE-2021-45105Medium· 5.9Apache Log4j2 versions 2.0-alpha1 through 2.16.0 (excluding 2.12.3 and 2.3.1) did not protect from uncontrolled recursion from self-referential lookups
CVE-2020-3478High· 8.1A vulnerability in the REST API of Cisco Enterprise NFV Infrastructure Software (NFVIS) could allow an authenticated, remote attacker to overwrite certain files that should be restricted on an affected device
CVE-2020-3577High· 7.4A vulnerability in the ingress packet processing path of Cisco Firepower Threat Defense (FTD) Software for interfaces that are configured either as Inline Pair or in Passive mode could allow an unauthenticated, adjacent attacker to cause…
CVE-2020-3317High· 7.5A vulnerability in the ssl_inspection component of Cisco Firepower Threat Defense (FTD) Software could allow an unauthenticated, remote attacker to crash Snort instances