Apache Opennlp

OOM Denial of Service via Unbounded Map Pre-Sizing in Apache OpenNLP SymSpellModelSerializer Versions Affected: - 3.0.0-M4 - 3.0.0-M5 (The opennlp-spellcheck extension was introduced in 3.0.0-M4. Releases 1.x and 2.x do not contain the affected code.) Description: The SymSpellModelSerializer.create() method reads two 32-bit signed integer count fields (unigramCount and bigramCount) from a binary SymSpell model stream and passes each value directly to LinkedHashMap.newLinkedHashMap() after validating only that it is non-negative. No upper bound is applied, so the count is fully attacker-controlled when the model file originates from an untrusted source. A crafted .bin model file in which either count field is set to Integer.MAX_VALUE (or any value large enough to exhaust the available heap) causes the map to be pre-sized to a capacity of 2^30 entries. The oversized backing array is allocated on the first put() into that map, requesting 4–8 GB depending on whether compressed oops are in effect, and the load fails with an OutOfMemoryError. Because the count fields sit immediately after a fixed-size header (magic, format version, three UTF strings, the configuration fields, and the edit-distance identifier) the attacker pays no meaningful size cost to weaponize a payload: a file of well under 100 bytes plus a single real entry is sufficient to crash a JVM that loads it. Any code path that deserializes a SymSpell model is affected, including SymSpellModels.deserialize(InputStream), SymSpellModels.fromBytes(byte[]), classpath model loading via SymSpellModelResolver.resolveByLanguage(String), the CorrectTextTool command-line tool, and model-archive loading through the registered ArtifactSerializer. The opennlp-spellcheck extension ships in the official OpenNLP binary distribution. The practical impact is denial of service against processes that load SymSpell model files from untrusted or semi-trusted origins. Mitigation: - 3.x users should upgrade to 3.0.0-M6. Note: The fix applies an upper bound to both count fields, checked before the map is pre-sized; counts that are negative or exceed the bound cause an IOException to be thrown and the read to fail fast with no large allocation. The bound is the existing AbstractModelReader.MAX_ENTRIES limit introduced earlie, which the current change promotes to public visibility so that serializers implementing their own binary format can share it. The default bound is 10,000,000, which is well above the entry counts of legitimate SymSpell dictionaries but far below any value that would threaten heap exhaustion. Deployments that legitimately need to load larger dictionaries can raise the limit at JVM startup by setting the OPENNLP_MAX_ENTRIES system property to the desired positive integer (e.g. -DOPENNLP_MAX_ENTRIES=50000000); invalid or non-positive values fall back to the default. Note that this property is shared with the model-reader limit and raising it relaxes both. Users who cannot upgrade immediately should treat all SymSpell .bin model files as untrusted input unless their provenance is verified, and should avoid loading models supplied by end users or fetched from third-party repositories without integrity checks.

Published: 2026-09-11T17:49:07.499Z
Updated: 2026-09-14T18:50:04.110Z

The two built-in name-finder patterns exposed by opennlp.tools.namefind.RegexNameFinderFactory - DEFAULT_REGEX_NAME_FINDER.EMAIL and DEFAULT_REGEX_NAME_FINDER.URL - contain ambiguous nested quantifiers. An application that obtains these finders through RegexNameFinderFactory.getDefaultRegexNameFinders(...) and then applies them to untrusted text through RegexNameFinder.find(String[]) or RegexNameFinder.find(String) can be driven into super-linear backtracking or into unbounded matcher recursion by a small crafted input. For the EMAIL pattern, a long run of local-part characters that is never followed by an @ forces the matcher to re-scan to end-of-input from every starting offset. Cost grows quadratically with input length: an input of approximately 32 KB consumes several seconds of CPU in a single find() call and returns no match, and each doubling of the input multiplies the cost roughly four-fold. For the URL pattern, the query-string sub-expression nests a capturing repetition inside an outer repetition. The JDK matcher recurses once per query token, so an input of approximately 4 KB containing many &-separated tokens exhausts the thread stack and causes java.lang.StackOverflowError to propagate out of find(), terminating the calling thread. On a thread created with a smaller stack (for example -Xss512k, typical of server worker pools) approximately 1 KB is sufficient. In both cases an attacker who can supply text for analysis can convert a single request into seconds to minutes of pinned CPU, or into an abrupt thread death, denying service to the embedding application. No authentication, special configuration, or model file is required beyond the application having selected one of the two built-in finders. This issue affects Apache OpenNLP: from 2.0.0 through 2.5.11; from 3.0.0-M1 through 3.0.0-M5. Users are recommended to upgrade to version 2.5.12, or to 3.0.0-M6 for users tracking the 3.0.0 milestone line, which fix the issue.

Published: 2026-09-11T17:50:53.639Z
Updated: 2026-09-11T21:07:10.400Z

Arbitrary Class Instantiation via XML Feature Generator Descriptor and Format Name in Apache OpenNLP Versions Affected: - before 2.5.10 - before 3.0.0-M5 Description: Three code paths in Apache OpenNLP load a class by its fully-qualified name via Class.forName() and invoke its no-arg constructor without any prior validation of the class name or its type.  The affected paths are: (1) GeneratorFactory, which reads the class attribute of generator elements in an XML feature generator descriptor; such descriptors are embedded as artifacts in model archives (e.g. TokenNameFinder and POSTagger models) and are parsed during model loading, so an attacker who can supply a crafted model archive controls the class name directly. (2) StreamFactoryRegistry.getFactory(Class, String), which falls back to interpreting an unregistered format name as the fully-qualified class name of an ObjectStreamFactory; this is exploitable in applications that pass untrusted format names (e.g. exposing the -format parameter of the command-line tooling to external input). (3) StringInterners, which instantiates the interner implementation named by the opennlp.interner.class system property; this value is normally deployer-controlled, so it is hardened as defense in depth rather than being independently attacker-reachable. Exploitation requires a class with attacker-useful side effects in its static initializer or no-arg constructor (JNDI lookup, outbound network I/O, filesystem access) to be present on the classpath, so this is not drop-in remote code execution. T Mitigation: Upgrade to a fixed release. The fix routes all three paths through ExtensionLoader.instantiateExtension(...), which consults a package-prefix allowlist before Class.forName() is invoked, so a disallowed class is never loaded, initialized, or constructed. Classes under the opennlp. prefix remain permitted by default. Deployments that load models referencing feature generator factories, object stream factories, or string interners outside opennlp.* must opt those packages in, either programmatically via ExtensionLoader.registerAllowedPackage(String) before the first model load, or by setting the OPENNLP_EXT_ALLOWED_PACKAGES system property to a comma-separated list of allowed package prefixes. Users who cannot upgrade immediately should ensure all model files and format names are sourced from trusted origins and should audit their classpath for classes with side-effecting static initializers or constructors.

Published: 2026-07-24T08:07:57.463Z
Updated: 2026-07-24T18:04:05.691Z

OOM Denial of Service via Unbounded Array Allocation in Apache OpenNLP AbstractModelReader  Versions Affected:  before 1.9.5 before 2.5.9 before 3.0.0-M3  Description: The AbstractModelReader methods getOutcomes(), getOutcomePatterns(), and getPredicates() each read a 32-bit signed integer count field from a binary model stream and pass that value directly to an array allocation (new String[numOutcomes], new int[numOCTypes][], new String[NUM_PREDS]) without validating that the value is non-negative or within a reasonable bound. The count is therefore fully attacker-controlled when the model file originates from an untrusted source. A crafted .bin model file in which any of these count fields is set to Integer.MAX_VALUE (or any value large enough to exhaust the available heap) triggers an OutOfMemoryError at the array allocation itself, before the corresponding label or pattern data is consumed from the stream. The error occurs very early in deserialization: for a GIS model, getOutcomes() is reached after only the model-type string, the correction constant, and the correction parameter have been read; so the attacker pays no meaningful size cost to weaponize a payload, and a single small file can crash a JVM that loads it. Any code path that deserializes a .bin model is affected, including direct use of GenericModelReader and any higher-level component that delegates to it during model load. The practical impact is denial of service against processes that load model files from untrusted or semi-trusted origins.   Mitigation: * 2.x users should upgrade to 2.5.9. * 3.x users should upgrade to 3.0.0-M3. Note: The fix introduces an upper bound on each of the three count fields, checked before array allocation; counts that are negative or exceed the bound cause an IllegalArgumentException to be thrown and the read to fail fast with no large allocation. The default bound is 10,000,000, which is well above the entry counts of legitimate OpenNLP models but far below any value that would threaten heap exhaustion. Deployments that legitimately need to load models with more entries than the default can raise the limit at JVM startup by setting the OPENNLP_MAX_ENTRIES system property to the desired positive integer (e.g. -DOPENNLP_MAX_ENTRIES=50000000); invalid or non-positive values fall back to the default. Users who cannot upgrade immediately should treat all .bin model files as untrusted input unless their provenance is verified, and should avoid loading models supplied by end users or fetched from third-party repositories without integrity checks.

Published: 2026-05-04T16:40:32.503Z
Updated: 2026-07-30T12:05:07.579Z

Arbitrary Class Instantiation via Model Manifest in Apache OpenNLP ExtensionLoader Versions Affected: before 1.9.5, before 2.5.9, before 3.0.0-M3 Description:  The ExtensionLoader.instantiateExtension(Class, String) method loads a class by its fully-qualified name via Class.forName() and invokes its no-arg constructor, with the class name sourced from the manifest.properties entry of a model archive. The existing isAssignableFrom check correctly rejects classes that are not subtypes of the expected extension interface (BaseToolFactory for factory=, ArtifactSerializer for serializer-class-*), but the check runs after Class.forName() has already loaded and initialized the named class. Class.forName() with default initialization semantics executes the target class's static initializer before returning, so an attacker who can supply a crafted model archive can cause the static initializer of any class on the classpath to run during model loading, regardless of whether that class passes the subsequent type check. Exploitation requires a class with attacker-useful side effects in its static initializer (for example, JNDI lookup, outbound network I/O, or filesystem access) to be present on the classpath, so this is not a drop-in remote code execution; however, the attack surface grows as third-party model distribution becomes more common (community model repositories, Hugging Face-style sharing), where users routinely load model files from origins they do not control. A secondary, narrower vector affects deployments that ship legitimate BaseToolFactory or ArtifactSerializer subclasses with side-effecting no-arg constructors: a malicious manifest can name such a class and force its constructor to run during model load. Mitigation:  * 2.x users should upgrade to 2.5.9. * 3.x users should upgrade to 3.0.0-M3. Note: The fix introduces a package-prefix allowlist that is consulted before Class.forName() is invoked, so the static initializer of a disallowed class is never executed. Classes under the opennlp. prefix remain permitted by default. Deployments that load models referencing factories or serializers outside opennlp.* must opt those packages in, either programmatically via ExtensionLoader.registerAllowedPackage(String) before the first model load, or by setting the OPENNLP_EXT_ALLOWED_PACKAGES system property to a comma-separated list of allowed package prefixes. Users who cannot upgrade immediately should ensure that all model files are sourced from trusted origins and should audit their classpath for classes with side-effecting static initializers or constructors, particularly any that perform JNDI lookups, network requests, or filesystem operations during class initialization.

Published: 2026-05-04T16:43:12.583Z
Updated: 2026-09-09T12:04:43.799Z