Please see the Configuring SSL Support page for more information.
Securing JAX-WS services
CXF supports WS-Security via the Apache WSS4J project. WSS4J provides an implementation of the following WS-Security standards:
- SOAP Message Security 1.1
- Username Token Profile 1.1
- X.509 Certificate Token Profile 1.1
- SAML Token Profile 1.1
- Kerberos Token Profile 1.1
- SOAP Messages with Attachments Profile 1.1
- Basic Security Profile 1.1
Please see the WS-Security page for more information.
CXF fully supports WS-SecurityPolicy, which allows you to configure WS-Security requirements for an endpoint using a WS-Policy annotation. This is the recommended way of configuring WS-Security. Policies can be added in a WSDL or else referenced via an annotation in code.
The WS-SecurityPolicy layer and the XML-Security layer in Apache CXF share a common set of security configuration tags from CXF 3.1.0. The Security Configuration page details these tags and values. There are also some addition configuration tags, that are only used for when security is configured via WS-SecurityPolicy, see the following page for more information.
CXF fully supports WS-SecureConveration, see the following page for more information.
CXF ships with a advanced SecurityTokenService (STS) implementation that can be used to issue (SAML) tokens for authentication. CXF also supports communicating with the STS using the WS-Trust specification. SSO is supported by caching the tokens on the client side. Please see the WS-Trust page for more information.
Securing JAX-RS services
JAX-RS XML Security
It is possible to secure XML based JAX-RS requests (and responses) using XML Signature and Encryption. See the JAX-RS XML Security page for more information.
See the JAX-RS SAML page on creating SAML Assertions and adding them to a JAX-RS request, as well as how to validate them on the receiving side.
SAML Web SSO
Please see this blog entry announcing the support for SAML Web SSO profile and the SAML Web SSO page for more information. CXF fully supports the SAML Web SSO profile on the service provider side. As of yet however, no IdP is available in CXF.
Apache CXF Fediz is a subproject of CXF. Fediz helps you to secure your web applications and delegates security enforcement to the underlying application server. With Fediz, authentication is externalized from your web application to an identity provider installed as a dedicated server component. The supported standard is WS-Federation Passive Requestor Profile. Fediz supports Claims Based Access Control beyond Role Based Access Control (RBAC).
Container or Spring Security managed authentication as well as the custom authentication are all the viable options used by CXF developers.
Starting from CXF 2.3.2 and 2.4.0 it is possible to use an org.apache.cxf.interceptor.security.JAASLoginInterceptor in order to authenticate a current user and populate a CXF SecurityContext.
The JAAS authenticator is configured with the name of the JAAS login context (the one usually specified in the JAAS configuration resource which the server is aware of). It is also configured with an optional "roleClassifier" property which is needed by the CXF SecurityContext in order to differentiate between user and role Principals. By default CXF will assume that role Principals are represented by javax.security.acl.Group instances.
In some cases objects representing a user principal and roles are implementing the same marker interface such as Principal. That can be handled like this:
In this case JAASLoginInterceptor will know that the roles are represented by a class whose simple name is RolePrincipal. Note that full class names are also supported.
Please see this page for the information about Spnego/Kerberos HTTPConduit client support.
Please check the following blog entries about WS-Security Kerberos support in CXF:
Please check the following page about Kerberos support in JAX-RS.
Container or Spring Security managed authorization as well as the custom authorization are all the viable options used by CXF developers.
CXF 2.3.2 and 2.4.0 introduce org.apache.cxf.interceptor.security.SimpleAuthorizingInterceptor and org.apache.cxf.interceptor.security.SecureAnnotationsInterceptor interceptors which can help with enforcing the authorization rules.
Controlling Large Request Payloads
Starting with CXF 2.7.4, CXF now requires use of a StAX parser that can provide fine grained control over the size of the incoming XML. The only parser that will currently work is Woodstox 4.2 or newer. The main reason is there are a series of DOS attacks that can only be prevented at the StAX parser level. There is a "org.apache.cxf.stax.allowInsecureParser" System Property that can be set to true to allow using an insecure parser, but that is HIGHLY not recommended and doing so would also now allow the settings described in this section.
CXF has several default settings that will prevent malicious XML from causing various DOS failures. You can override the default values if you know you will have incoming XML that will exceed these limits. These settings can be set as Bus level properties, endpoint level properties, or even per request via an interceptor.
Maximum number of child elements for a given parent element
Maximum depth of an element
Maximum number of attributes on a single element
Maximum size of a single attribute
Maximum size of an elements text value
Maximum total number of elements in the XML document
Maximum total number of characters parsed by the parser
XML - CXF versions prior to 2.7.4
Endpoints expecting XML payloads may get DepthRestrictingInterceptor registered and configured in order to control the limits a given XML payload may not exceed. This can be useful in a variety of cases in order to protect against massive payloads which can potentially cause the denial-of-service situation or simply slow the service down a lot.
The complete number of XML elements, the number of immediate children of a given XML element may contain and the stack depth of the payload can be restricted, for example:
When one of the limits is reached, the error is returned. JAX-WS consumers will receive 500, JAX-RS/HTTP consumers: 413.
The following system properties can also be set up for JAX-WS endpoints: "org.apache.cxf.staxutils.innerElementCountThreshold" and "org.apache.cxf.staxutils.innerElementLevelThreshold".
Please check this section for the additional information on how JAX-RS JAXB-based providers can be configured.
The "org.apache.cxf.io.CachedOutputStream.MaxSize" system property or "attachment-max-size" per-endpoint contextual property can be used to control the size of large attachments. When the limits is reached, the error is returned. JAX-WS consumers will receive 500, JAX-RS/HTTP consumers: 413.
Large data stream caching
A large stream based message or data will be cached in a temporary file. In default, this caching occurs at data size larger than 64K bytes and a temporary file is written in the system's temporary directory. You can change this behavior and other properties of the caching feature by explicitly setting the following properties.
To change the default behavior for the entire system, you can set the following system properties.
The threshold value in bytes to switch from memory to file caching
The data size in bytes to limit the maximum data size to be cached
The directory name for storing the temporary files
The cipher transformation name for encryptiing the cached content
To change the default behavior for a specific bus, you can set the corresponding bus.io.CachedOutputStream properties (e.g., bus.io.CachedOutputStream.Threshold for org.apache.cxf.io.CachedOutputStream.Threshold).
The encryption option, which is available from CXF 2.6.4 and 2.7.1, uses a symmetric encryption using a generated key and it can be used to protect the cached content from unauthorized access. To enable encryption, the CipherTransformation property can be set to the name of an appropriate stream or 8-bit block cipher transformation (e.g., RC4, AES/CTR/NoPadding, etc) that is supported by the environment. However, it is noted that enabling the encryption will result in an increased processing time and it is therefore recommended only in specific use cases where other means to protect the cached content is unavailable.