Apache OpenNLP versions before 2.5.9 and 3.0.0-M3 contain an XXE vulnerability in the DictionaryEntryPersistor class. This class initializes a SAXParserFactory without enabling FEATURE_SECURE_PROCESSING or disabling DTD processing, leaving external entity resolution and DOCTYPE declarations enabled. When the create(InputStream, EntryInserter) method is called, an attacker supplying a crafted dictionary XML file with malicious DOCTYPE declarations can exploit this to perform local file disclosure or server-side request forgery during SAX parsing. This vulnerability affects the public Dictionary(InputStream) constructor, which is the documented API for loading user dictionaries, making untrusted input a realistic attack vector. The vulnerability is inconsistent with other XML parsing in the project, which correctly disables these features. The vendor recommends upgrading to versions 2.5.9 or 3.0.0-M3 to remediate the issue.

An attacker able to supply a crafted dictionary XML file to the vulnerable Apache OpenNLP versions can exploit this XXE vulnerability to read local files on the server or cause server-side request forgery. This can lead to unauthorized disclosure of sensitive information or interaction with internal network resources. The impact is limited to scenarios where untrusted dictionary files are processed by the vulnerable API.

A fixed version is available: upgrade to Apache OpenNLP 2.5.9 or 3.0.0-M3. If upgrading immediately is not possible, ensure all dictionary files are sourced from trusted origins. Additionally, implement input validation to reject any XML containing DOCTYPE declarations before it is parsed by the Dictionary(InputStream) constructor. These mitigations reduce the risk of exploitation until an upgrade can be performed.