v0.12.6

Compare Source

This patch release:

• Ensures that after successful JWS signature verification, an application-configured Base64Url Decoder output is
  used to construct a Jws instance (instead of JJWT's default decoder). See
  Issue 947.
• Fixes a decompression memory leak in concurrent/multi-threaded environments introduced in 0.12.0 when decompressing JWTs with a zip header of GZIP. See Issue 949.
• Upgrades BouncyCastle to 1.78 via PR 941.

v0.12.5

Compare Source

This patch release:

• Ensures that builders' NestedCollection changes are applied to the collection immediately as mutation methods are called, no longer
  requiring application developers to call .and() to 'commit' or apply a change. For example, prior to this release,
  the following code did not apply changes:

  JwtBuilder builder = Jwts.builder();
  builder.audience().add("an-audience"); // no .and() call
  builder.compact(); // would not keep 'an-audience'

  Now this code works as expected and all other NestedCollection instances like it apply changes immediately (e.g. when calling
  .add(value)).

  However, standard fluent builder chains are still recommended for readability when feasible, e.g.

  Jwts.builder()
      .audience().add("an-audience").and() // allows fluent chaining
      .subject("Joe")
      // etc...
      .compact()

  See Issue 916.

v0.12.4

Compare Source

This patch release includes various changes listed below.

Jackson Default Parsing Behavior

This release makes two behavioral changes to JJWT's default Jackson ObjectMapper parsing settings:

1. In the interest of having stronger standards to reject potentially malformed/malicious/accidental JSON that could
   have undesirable effects on an application, JJWT's default ObjectMapper is now configured to explicitly reject/fail
   parsing JSON (JWT headers and/or Claims) if/when that JSON contains duplicate JSON member names.

   For example, now the following JSON, if parsed, would fail (be rejected) by default:

   {
     "hello": "world",
     "thisWillFail": 42,
     "thisWillFail": "test"
   }

   Technically, the JWT RFCs do allow duplicate named fields as long as the last parsed member is the one used
   (see JWS RFC 7515, Section 4), so this is allowed.
   However, because JWTs often reflect security concepts, it's usually better to be defensive and reject these
   unexpected scenarios by default. The RFC later supports this position/preference in
   Section 10.12:

   Ambiguous and potentially exploitable situations
   could arise if the JSON parser used does not enforce the uniqueness
   of member names or returns an unpredictable value for duplicate
   member names.
   

   Finally, this is just a default, and the RFC does indeed allow duplicate member names if the last value is used,
   so applications that require duplicates to be allowed can simply configure their own ObjectMapper and use
   that with JJWT instead of assuming this (new) JJWT default. See
   Issue #​877 for more.

2. If using JJWT's support to use Jackson to parse
   Custom Claim Types (for example, a Claim that should be
   unmarshalled into a POJO), and the JSON for that POJO contained a member that is not represented in the specified
   class, Jackson would fail parsing by default. Because POJOs and JSON data models can sometimes be out of sync
   due to different class versions, the default behavior has been changed to ignore these unknown JSON members instead
   of failing (i.e. the ObjectMapper's DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES is now set to false)
   by default.

   Again, if you prefer the stricter behavior of rejecting JSON with extra or unknown properties, you can configure
   true on your own ObjectMapper instance and use that instance with the Jwts.parser() builder.

Additional Changes

This release also:

• Fixes a thread-safety issue when using java.util.ServiceLoader to dynamically lookup/instantiate pluggable
  implementations of JJWT interfaces (e.g. JSON parsers, etc). See
  Issue #​873 and its documented fix in
  PR #​893.
• Ensures Android environments and older org.json library usages can parse JSON from a JwtBuilder-provided
  java.io.Reader instance. Issue 882.
• Ensures a single string aud (Audience) claim is retained (without converting it to a Set) when copying/applying a
  source Claims instance to a destination Claims builder. Issue 890.
• Ensures P-256, P-384 and P-521 Elliptic Curve JWKs zero-pad their field element (x, y, and d) byte array values
  if necessary before Base64Url-encoding per RFC 7518, Sections
  6.2.1.2,
  6.2.1.3, and
  6.2.2.1, respectively.
  Issue 901.
• Ensures that Secret JWKs for HMAC-SHA algorithms with k sizes larger than the algorithm minimum can
  be parsed/used as expected. See Issue #​905
• Ensures there is an upper bound (maximum) iterations enforced for PBES2 decryption to help mitigate potential DoS
  attacks. Many thanks to Jingcheng Yang and Jianjun Chen from Sichuan University and Zhongguancun Lab for their
  work on this. See PR 911.
• Fixes various typos in documentation and JavaDoc. Thanks to those contributing pull requests for these!

v0.12.3

Compare Source

This patch release:

• Upgrades the org.json dependency to 20231013 to address that library's
  CVE-2023-5072 vulnerability.
• (Re-)enables empty values for custom claims, which was the behavior in <= 0.11.5.
  Issue 858.

v0.12.2

Compare Source

This is a follow-up release to finalize the work in 0.12.1 that tried to fix a reflection scope problem
on >= JDK 17. The 0.12.1 fix worked, but only if the importing project or application did not have its own
module-info.java file.

This release removes that reflection code entirely in favor of a JJWT-native implementation, eliminating JPMS
module (scope) problems on >= JDK 17. As such, --add-opens flags are no longer required to use JJWT.

The fix has been tested up through JDK 21 in a separate application environment (out of JJWT's codebase) to assert
expected functionality in a 'clean room' environment in a project both with and without module-info.java usage.

v0.12.1

Compare Source

Enabled reflective access on JDK 17+ to java.io.ByteArrayInputStream and sun.security.util.KeyUtil for
jjwt-impl.jar

v0.12.0

Compare Source

This is a big release! JJWT now fully supports Encrypted JSON Web Tokens (JWE), JSON Web Keys (JWK) and more! See the
sections below enumerating all new features as well as important notes on breaking changes or backwards-incompatible
changes made in preparation for the upcoming 1.0 release.

Because breaking changes are being introduced, it is strongly recommended to wait until the upcoming 1.0 release
where you can address breaking changes one time only.

Those that need immediate JWE encryption and JWK key support
however will likely want to upgrade now and deal with the smaller subset of breaking changes in the 1.0 release.

Simplified Starter Jar

Those upgrading to new modular JJWT versions from old single-jar versions will transparently obtain everything
they need in their Maven, Gradle or Android projects.

JJWT's early releases had one and only one .jar: jjwt.jar. Later releases moved to a modular design with 'api' and
'impl' jars including 'plugin' jars for Jackson, GSON, org.json, etc. Some users upgrading from the earlier single
jar to JJWT's later versions have been frustrated by being forced to learn how to configure the more modular .jars.

This release re-introduces the jjwt.jar artifact again, but this time it is simply an empty .jar with Maven
metadata that will automatically transitively download the following into a project, retaining the old single-jar
behavior:

• jjwt-api.jar
• jjwt-impl.jar
• jjwt-jackson.jar

Naturally, developers are still encouraged to configure the modular .jars as described in JJWT's documentation for
greater control and to enable their preferred JSON parser, but this stop-gap should help those unaware when upgrading.

JSON Web Encryption (JWE) Support!

This has been a long-awaited feature for JJWT, years in the making, and it is quite extensive - so many encryption
algorithms and key management algorithms are defined by the JWA specification, and new API concepts had to be
introduced for all of them, as well as extensive testing with RFC-defined test vectors. The wait is over!
All JWA-defined encryption algorithms and key management algorithms are fully implemented and supported and
available immediately. For example:

AeadAlgorithm enc = Jwts.ENC.A256GCM;
SecretKey key = enc.key().build();
String compact = Jwts.builder().setSubject("Joe").encryptWith(key, enc).compact();

Jwe<Claims> jwe = Jwts.parser().decryptWith(key).build().parseEncryptedClaims(compact);

Many other RSA and Elliptic Curve examples are in the full README documentation.

JSON Web Key (JWK) Support!

Representing cryptographic keys - SecretKeys, RSA Public and Private Keys, Elliptic Curve Public and
Private keys - as fully encoded JSON objects according to the JWK specification - is now fully implemented and
supported. The new Jwks utility class exists to create JWK builders and parsers as desired. For example:

SecretKey key = Jwts.SIG.HS256.key().build();
SecretJwk jwk = Jwks.builder().forKey(key).build();
assert key.equals(jwk.toKey());

// or if receiving a JWK string:
Jwk<?> parsedJwk = Jwks.parser().build().parse(jwkString);
assert jwk.equals(parsedJwk);
assert key.equals(parsedJwk.toKey());

Many JJWT users won't need to use JWKs explicitly, but some JWA Key Management Algorithms (and lots of RFC test
vectors) utilize JWKs when transmitting JWEs. As this was required by JWE, it is now implemented in full for
JWE use as well as general-purpose JWK support.

JWK Thumbprint and JWK Thumbprint URI support

The JWK Thumbprint and
JWK Thumbprint URI RFC specifications are now fully supported. Please
see the README.md file's corresponding named sections for both for full documentation and usage examples.

JWS Unencoded Payload Option (b64) support

The JSON Web Signature (JWS) Unencoded Payload Option RFC specification
is now fully supported. Please see the README.md corresponding named section for documentation and usage examples.

Better PKCS11 and Hardware Security Module (HSM) support

Previous versions of JJWT enforced that Private Keys implemented the RSAKey and ECKey interfaces to enforce key
length requirements. With this release, JJWT will still perform those checks when those data types are available,
but if not, as is common with keys from PKCS11 and HSM KeyStores, JJWT will still allow those Keys to be used,
expecting the underlying Security Provider to enforce any key requirements. This should reduce or eliminate any
custom code previously written to extend JJWT to use keys from those KeyStores or Providers.

Additionally, PKCS11/HSM tests using SoftHSMv2 are run on every build with
every JWS MAC and Signature algorithm and every JWE Key algorithm to ensure continued stable support with
Android and Sun PKCS11 implementations and spec-compliant Hardware Security Modules that use the PKCS11 interface
(such as YubiKey, etc.)

Custom Signature Algorithms

The io.jsonwebtoken.SignatureAlgorithm enum has been deprecated in favor of new
io.jsonwebtoken.security.SecureDigestAlgorithm, io.jsonwebtoken.security.MacAlgorithm, and
io.jsonwebtoken.security.SignatureAlgorithm interfaces to allow custom algorithm implementations. The new nested
Jwts.SIG static inner class is a registry of all standard JWS algorithms as expected, exactly like the
old enum. This change was made because enums are a static concept by design and cannot
support custom values: those who wanted to use custom signature algorithms could not do so until now. The new
interfaces now allow anyone to plug in and support custom algorithms with JJWT as desired.

KeyBuilder and KeyPairBuilder

Because the io.jsonwebtoken.security.Keys#secretKeyFor and io.jsonwebtoken.security.Keys#keyPairFor methods
accepted the now-deprecated io.jsonwebtoken.SignatureAlgorithm enum, they have also been deprecated in favor of
calling new key() or keyPair() builder methods on MacAlgorithm and SignatureAlgorithm instances directly.
For example:

SecretKey key = Jwts.SIG.HS256.key().build();
KeyPair pair = Jwts.SIG.RS256.keyPair().build();

The builders allow for customization of the JCA Provider and SecureRandom during Key or KeyPair generation if desired, whereas
the old enum-based static utility methods did not.

Preparation for 1.0

Now that the JWE and JWK specifications are implemented, only a few things remain for JJWT to be considered at
version 1.0. We have been waiting to apply the 1.0 release version number until the entire set of JWT specifications
are fully supported and we drop JDK 7 support (to allow users to use JDK 8 APIs). To that end, we have had to
deprecate some concepts, or in some cases, completely break backwards compatibility to ensure the transition to
1.0 (and JDK 8 APIs) are possible. Most backwards-incompatible changes are listed in the next section below.

Backwards Compatibility Breaking Changes, Warnings and Deprecations

• io.jsonwebtoken.Jwt's getBody() method has been deprecated in favor of a new getPayload() method to
  reflect correct JWT specification nomenclature/taxonomy.

• io.jsonwebtoken.Jws's getSignature() method has been deprecated in favor of a new getDigest() method to
  support expected congruent behavior with Jwe instances (both have digests).

• io.jsonwebtoken.JwtParser's parseContentJwt, parseClaimsJwt, parseContentJws, and parseClaimsJws methods
  have been deprecated in favor of more intuitive respective parseUnsecuredContent, parseUnsecuredClaims,
  parseSignedContent and parseSignedClaims methods.

• io.jsonwebtoken.CompressionCodec is now deprecated in favor of the new io.jsonwebtoken.io.CompressionAlgorithm
  interface. This is to guarantee API congruence with all other JWT-identifiable algorithm IDs that can be set as a
  header value.

• io.jsonwebtoken.CompressionCodecResolver has been deprecated in favor of the new
  JwtParserBuilder#addCompressionAlgorithms method.

Breaking Changes

• io.jsonwebtoken.Claims and io.jsonwebtoken.Header instances are now immutable to enhance security and thread
  safety. Creation and mutation are supported with newly introduced ClaimsBuilder and HeaderBuilder concepts.
  Even though mutation methods have migrated, there are a couple that have been removed entirely:

  • io.jsonwebtoken.JwsHeader#setAlgorithm has been removed - the JwtBuilder will always set the appropriate
    alg header automatically based on builder state.
  • io.jsonwebtoken.Header#setCompressionAlgorithm has been removed - the JwtBuilder will always set the appropriate
    zip header automatically based on builder state.

• io.jsonwebtoken.Jwts's header(Map), jwsHeader() and jwsHeader(Map) methods have been removed in favor
  of the new header() method that returns a HeaderBuilder to support method chaining and dynamic Header type
  creation. The HeaderBuilder will dynamically create a Header, JwsHeader or JweHeader automatically based on
  builder state.

• Similarly, io.jsonwebtoken.Jwts's claims() static method has been changed to return a ClaimsBuilder instead
  of a Claims instance.

• JWTs that do not contain JSON Claims now have a payload type of byte[] instead of String (that is,
  Jwt<byte[]> instead of Jwt<String>). This is because JWTs, especially when used with the
  cty (Content Type) header, are capable of handling any type of payload, not just Strings. The previous JJWT
  releases didn't account for this, and now the API accurately reflects the JWT RFC specification payload
  capabilities. Additionally, the name of plaintext has been changed to content in method names and JavaDoc to
  reflect this taxonomy. This change has impacted the following JJWT APIs:

  • The JwtBuilder's setPayload(String) method has been deprecated in favor of two new methods:

    • setContent(byte[]), and
    • setContent(byte[], String contentType)

    These new methods allow any kind of content
    within a JWT, not just Strings. The existing setPayload(String) method implementation has been changed to
    delegate to this new setContent(byte[]) method with the argument's UTF-8 bytes, for example
    setContent(payloadString.getBytes(StandardCharsets.UTF_8)).

  • The JwtParser's Jwt<Header, String> parsePlaintextJwt(String plaintextJwt) and
    Jws<String> parsePlaintextJws(String plaintextJws) methods have been changed to
    Jwt<Header, byte[]> parseContentJwt(String plaintextJwt) and
    Jws<byte[]> parseContentJws(String plaintextJws) respectively.

  • JwtHandler's onPlaintextJwt(String) and onPlaintextJws(String) methods have been changed to
    onContentJwt(byte[]) and onContentJws(byte[]) respectively.

  • io.jsonwebtoken.JwtHandlerAdapter has been changed to reflect the above-mentioned name and String-to-byte[]
    argument changes, as well adding the abstract modifier. This class was never intended
    to be instantiated directly, and is provided for subclassing only. The missing modifier has been added to ensure
    the class is used as it had always been intended.

  • io.jsonwebtoken.SigningKeyResolver's resolveSigningKey(JwsHeader, String) method has been changed to
    resolveSigningKey(JwsHeader, byte[]).

• io.jsonwebtoken.JwtParser is now immutable. All mutation/modification methods (setters, etc) deprecated 4 years
  ago have been removed. All parser configuration requires using the JwtParserBuilder.

• Similarly, io.jsonwebtoken.Jwts's parser() method deprecated 4 years ago has been changed to now return a
  JwtParserBuilder instead of a direct JwtParser instance. The previous Jwts.parserBuilder() method has been
  removed as it is now redundant.

• The JwtParserBuilder no longer supports PrivateKeys for signature verification. This was an old
  legacy behavior scheduled for removal years ago, and that change is now complete. For various cryptographic/security
  reasons, asymmetric public/private key signatures should always be created with PrivateKeys and verified with
  PublicKeys.

• io.jsonwebtoken.CompressionCodec implementations are no longer discoverable via java.util.ServiceLoader due to
  runtime performance problems with the JDK's ServiceLoader implementation per
  https://github.com/jwtk/jjwt/issues/648/648. Custom implementations should be made available to the JwtParser via
  the new JwtParserBuilder#addCompressionAlgorithms method.

• Prior to this release, if there was a serialization problem when serializing the JWT Header, an IllegalStateException
  was thrown. If there was a problem when serializing the JWT claims, an IllegalArgumentException was
  thrown. This has been changed up to ensure consistency: any serialization error with either headers or claims
  will now throw a io.jsonwebtoken.io.SerializationException.

• Parsing of unsecured JWTs (alg header of none) are now disabled by default as mandated by
  RFC 7518, Section 3.6. If you require parsing of
  unsecured JWTs, you must call the JwtParserBuilder#enableUnsecured() method, but note the security
  implications mentioned in that method's JavaDoc before doing so.

• io.jsonwebtoken.gson.io.GsonSerializer now requires Gson instances that have a registered
  GsonSupplierSerializer type adapter, for example:

  new GsonBuilder()
    .registerTypeHierarchyAdapter(io.jsonwebtoken.lang.Supplier.class, GsonSupplierSerializer.INSTANCE)    
    .disableHtmlEscaping().create();

  This is to ensure JWKs have toString() and application log safety (do not print secure material), but still
  serialize to JSON correctly.

• io.jsonwebtoken.InvalidClaimException and its two subclasses (IncorrectClaimException and MissingClaimException)
  were previously mutable, allowing the corresponding claim name and claim value to be set on the exception after
  creation. These should have always been immutable without those setters (just getters), and this was a previous
  implementation oversight. This release has ensured they are immutable without the setters.