Oracle® Containers for J2EE Security Guide
10g Release 3 (10.1.3) B14429-01 |
|
Previous |
Next |
This chapter discusses security issues affecting EJBs, covering the following topics:
Note that beginning with the OC4J 10.1.3 implementation, the EJB container supports the OracleAS JAAS Provider.
See Also:
|
There are two JNDI properties that are specific to security. You can either set these within the jndi.properties
file or within your client implementation.
If setting the JNDI properties within the jndi.properties
file, set the properties as follows. Make sure that this jndi.properties
file is accessible from the classpath.
When you access EJBs in a remote container, you must pass valid credentials to this container. standalone clients define their credentials in the jndi.properties
file deployed with the client's code.
java.naming.security.principal=username java.naming.security.credentials=password
JavaBeans running within the container pass their credentials within the InitialContext
instance, which is created to look up the remote EJBs.
For example, to pass JNDI security properties within the Hashtable
environment:
Hashtable env = new Hashtable();
env.put("java.naming.provider.url", "ormi://myhost/ejbsamples");
env.put("java.naming.factory.initial",
"oracle.j2ee.naming.ApplicationClientInitialContextFactory");
env.put(Context.SECURITY_PRINCIPAL, "guest");
env.put(Context.SECURITY_CREDENTIALS, "welcome");
Context ic = new InitialContext (env);
Object homeObject = ic.lookup("java:comp/env/employeeBean");
// Narrow the reference to a TemplateHome.
EmployeeHome empHome =
(EmployeeHome) PortableRemoteObject.narrow(homeObject, EmployeeHome.class);
Note: ApplicationClientInitialContextFactory is in the file oc4jclient.jar .
|
EJB security involves two realms: granting permissions if you download into a browser, and configuring your application for authentication and authorization. This section covers the following:
If you download the EJB application as a client where the security manager is active, you must grant the following permissions before you can execute:
permission java.net.SocketPermission "*:*", "connect,resolve"; permission java.lang.RuntimePermission "createClassLoader"; permission java.lang.RuntimePermission "getClassLoader"; permission java.util.PropertyPermission "*", "read"; permission java.util.PropertyPermission "LoadBalanceOnLookup", "read,write";
You can define security constraints and J2EE security roles in the EJB deployment descriptor to protect your EJB methods (or in the orion-application.xml
descriptor for the overall J2EE application). These J2EE security roles must be mapped to deployment users and roles in the OracleAS JAAS Provider. This security role mapping can be accomplished through Application Server Control during deployment, as described in "Specifying Security Role Mapping through Application Server Control".
For authentication and authorization, this section focuses on XML configuration within the EJB deployment descriptors. EJB authorization is specified within the EJB and OC4J-specific deployment descriptors. You can manage the authorization piece of your security within the deployment descriptors, as follows:
The EJB deployment descriptor describes access rules using logical roles.
The OC4J-specific deployment descriptor maps the logical roles to deployment users and roles, made available to OC4J through the OracleAS JAAS Provider.
Users and roles are identities known by the container. Roles are the logical identities each application uses to indicate access rights to its different objects. The user name / password pairs can be digital certificates and, in the case of SSL, private key pairs.
Thus, the definition and mapping of roles is demonstrated in Figure 14-1.
Figure 14-1 End-to-End Security Role Configuration
Defining users and roles are discussed in the following sections:
As shown in Figure 14-2, you can use a logical name for a role within your bean implementation, and map this logical name to the correct security role or user. The mapping of the logical name to a database role is specified in the OC4J-specific deployment descriptor. See "Mapping Logical Roles to Users and Roles" for more information.
If you use a logical name for a database role within your bean implementation for methods such as isCallerInRole()
, you can map the logical name to an actual database role by doing the following:
Declare the logical name within the <enterprise-beans>
section in a <security-role-ref>
element. For example, to define a role used within the purchase order example, you may have checked within the bean implementation to see if the caller had authorization to sign a purchase order. Thus, the caller would have to be signed in under a correct role. For the bean to not require awareness of database roles, you can check isCallerInRole()
on a logical name, such as POMgr
, because only purchase order managers can sign off on the order. Thus, you would specify the logical security role, POMgr
, in the <role-name>
subelement of a <security-role-ref>
element within the <enterprise-beans>
section, as follows:
<enterprise-beans> ... <security-role-ref> <role-name>POMgr</role-name> <role-link>myMgr</role-link> </security-role-ref> ... </enterprise-beans>
The <role-link>
setting within the <security-role-ref>
element can be the actual database role, which is defined further within the <assembly-descriptor>
section. Alternatively, it can be another logical name, which is still defined more in the <assembly-descriptor>
section and is mapped to an actual database role within the Oracle-specific deployment descriptor.
Note: The<security-role-ref> element is required only when you use security context methods within your bean.
|
Define the role and the methods that it applies to. In the purchase order example, any method executed within the PurchaseOrder
bean must have authorized itself as myMgr
. Note that PurchaseOrder
is the name declared in the <ejb-name>
element, a subelement of the <session>
or <entity>
element.
Thus, the following defines the role as myMgr
, the EJB as PurchaseOrder
, and all methods by denoting the "*
" symbol.
Note: ThemyMgr role in the <security-role> element is the same as the <role-link> setting within the <enterprise-beans> section. This ties the logical name of POMgr to the myMgr definition.
|
<assembly-descriptor> <security-role> <description>Role needed purchase order authorization</description> <role-name>myMgr</role-name> </security-role> <method-permission> <role-name>myMgr</role-name> <method> <ejb-name>PurchaseOrder</ejb-name> <method-name>*</method-name> </method> </method-permission> ... </assembly-descriptor>
After performing both steps, you can refer to POMgr
within the bean implementation, and the container translates POMgr
to myMgr
.
The <method>
subelement of <method-permission>
is used to specify the security role for one or more methods within an interface or implementation. According to the EJB specification, this definition can be of one of the following forms:
Defining all methods within a bean by specifying the bean name and using the "*
" character to denote all methods within the bean, as follows:
<method-permission> <role-name>myMgr</role-name> <method> <ejb-name>EJBNAME</ejb-name> <method-name>*</method-name> </method> </method-permission>
Defining a specific method that is uniquely identified within the bean. Use the appropriate interface name and method name, as follows:
<method-permission> <role-name>myMgr</role-name> <method> <ejb-name>myBean</ejb-name> <method-name>myMethodInMyBean</method-name> </method> </method-permission>
Note: If there are multiple methods with the same overloaded name, the element of this style refers to all the methods with the overloaded name. |
Defining a method with a specific signature among many overloaded versions, as follows:
<method-permission> <role-name>myMgr</role-name> <method> <ejb-name>myBean</ejb-name> <method-name>myMethod</method-name> <method-params> <method-param>javax.lang.String</method-param> <method-param>javax.lang.String</method-param> </method-params> </method> </method-permission>
The parameters are the fully-qualified Java types of the input parameters of the method. If the method has no input arguments, the <method-params
> element contains no subelements.
If you want certain methods to not be checked for security roles, you define these methods as unchecked, as follows:
<method-permission> <unchecked/> <method> <ejb-name>EJBNAME</ejb-name> <method-name>*</method-name> </method> </method-permission>
Instead of defining a <role-name>
element, you define an <unchecked/>
element. When executing any methods in the EJBNAME
bean, the container does not check for security. Unchecked methods always override any other role definitions.
You can specify that all methods of an EJB execute under a specific identity. That is, the container does not check different roles for permission to run specific methods; instead, the container executes all of the EJB methods under the specified security identity. You can specify a particular role or the caller identity as the security identity.
Specify the "run-as" security identity in the <security-identity>
element, which is contained in the <enterprise-beans>
section. The following XML demonstrates that POMgr
is the role under which all the entity bean methods execute:
<enterprise-beans> <entity> ... <security-identity> <run-as> <role-name>POMgr</role-name> </run-as> </security-identity> ... </entity> </enterprise-beans>
Alternatively, the following XML example demonstrates how to specify that all methods of the bean execute under the identity of the caller:
<enterprise-beans> <entity> ... <security-identity> <use-caller-identity/> </security-identity> ... </entity> </enterprise-beans>
As noted earlier, you can define security constraints and J2EE security roles in the EJB deployment descriptor to protect your EJB methods. These J2EE security roles must be mapped to deployment users and roles in the OracleAS JAAS Provider. This security role mapping can be accomplished through Application Server Control during deployment, as described in "Specifying Security Role Mapping through Application Server Control".
Mappings are reflected in <security-role-mapping>
settings in Oracle-specific descriptors, as shown in the examples that follow.
See Also:
|
Example 14-1 Mapping Logical Role to Actual Role
While we recommend that you use Application Server Control for role mapping, this example provides reference information for the resulting configuration in orion-ejb-jar.xml
when you map the logical role POMGR
to the managers
role. Any user that can log in as part of this role is considered to have the POMGR
role, and so can execute the methods of PurchaseOrderBean
.
<security-role-mapping name="POMGR"> <group name="managers" /> </security-role-mapping>
For mapping to a specific user:
<security-role-mapping name="POMGR"> <user name="guest" /> </security-role-mapping>
For mapping to a specific user within a specific role:
<security-role-mapping name="POMGR"> <group name="managers" /> <user name="guest" /> </security-role-mapping>
As shown in Figure 14-3, the logical role name for POMGR
defined in the EJB deployment descriptor is mapped to managers
within the OC4J-specific deployment descriptor, in the <security-role-mapping>
element.
Notice that the <role-name>
setting in the EJB deployment descriptor is the same as for the name
attribute in the <security-role-mapping>
element in the OC4J-specific deployment descriptor. This is what identifies the mapping.
If any methods have not been associated with a role mapping, they are mapped to the default security role through the <default-method-access>
element in the orion-ejb-jar.xml
file. The following is the automatic mapping for any unsecured methods:
<default-method-access> <security-role-mapping name="<default-ejb-caller-role>" impliesAll="true" /> </default-method-access>
The default role is <default-ejb-caller-role>
and is defined in the name
attribute. You can replace this string with any name for the default role. The impliesAll
attribute indicates whether any security role checking occurs for these methods. This attribute defaults to "true
", which states that no security role checking occurs for these methods. If you set this attribute to "false
", the container will check for this default role on these methods.
If the impliesAll
attribute is "false
", you must map the default role defined in the name
attribute to a deployment user or role through the <user>
and <group>
elements. The following example shows how all methods not associated with a method permission are mapped to the "others
" role.
<default-method-access> <security-role-mapping name="default-role" impliesAll="false" /> <group name="others" /> </security-role-mapping> </default-method-access>
When you access EJBs in a remote container, you must pass valid credentials to this container.
Standalone clients define their credentials in the jndi.properties
file deployed with the EAR file.
Servlets or JavaBeans running within the container pass their credentials within the InitialContext
, which is created to look up the remote EJBs.
Indicate the user name (principal) and password (credentials) to use when looking up remote EJBs in the jndi.properties
file.
For example, if you want to access remote EJBs as POMGR/welcome
, define the following properties. The java.naming.factory.initial
setting indicates that you will use the Oracle JNDI implementation:
java.naming.security.principal=POMGR java.naming.security.credentials=welcome java.naming.factory.initial= oracle.j2ee.naming.ApplicationClientInitialContextFactory java.naming.provider.url=ormi://myhost/ejbsamples
In your application program, authenticate and access the remote EJBs as shown in the following example:
InitialContext ic = new InitialContext(); CustomerHome = (CustomerHome)ic.lookup("java:comp/env/purchaseOrderBean");
To access remote EJBs from a servlet or JavaBean, pass the credentials in the InitialContext
object as follows:
Hashtable env = new Hashtable(); env.put("java.naming.provider.url", "ormi://myhost/ejbsamples"); env.put("java.naming.factory.initial", "oracle.j2ee.naming.ApplicationClientInitialContextFactory"); env.put(Context.SECURITY_PRINCIPAL, "POMGR"); env.put(Context.SECURITY_CREDENTIALS, "welcome"); Context ic = new InitialContext (env); CustomerHome = (CustomerHome)ic.lookup("java:comp/env/purchaseOrderBean");
Anonymous EJB lookup is a mode you may consider, presumably only during early development or very special circumstances. In this mode, you do not specify the principal and credential when creating the InitialContext
, and therefore do not have to specify a principal or credential to remotely access EJBs. Your jndi.properties
file would look like this:
java.naming.factory.initial= oracle.j2ee.naming.ApplicationClientInitialContextFactory java.naming.provider.url=ormi://localhost:23791/ejb30slsb java.naming.security.principal= java.naming.security.credentials=
Important: Oracle strongly discourages this practice, except in special circumstances, as it leaves EJBs completely unsecured. |
You can enable this mode as follows:
Confirm that the anonymous
user is configured in system-jazn-data.xml
, and that this user is activated, as described in "Predefined OC4J Accounts".
Also in system-jazn-data.xml
, under the appropriate realm, assign the anonymous
user to a role that has been granted RMI permissions, as described in the next section, "Permitting EJB RMI Client Access". For example, assuming the users
role is granted RMI permissions:
<jazn-data> ... <jazn-realm> <realm> <name>myrealm</name> ... <roles> <role> <name>users</name> <members> <member> <type>user</type> <name>anonymous</name> </member> </members> </role> ... </roles> ... </realm> ... </jazn-realm> ... </jazn-data>
Give the role (users
in this example) appropriate namespace access so that it can execute read and write operations on the EJB. Use configuration such as the following in the orion-application.xml
file for the application:
<orion-application> ... <namespace-access> <read-access> <namespace-resource root=""> <security-role-mapping name="<jndi-user-role>"> <group name="administrators" /> <group name="users" /> </security-role-mapping> </namespace-resource> </read-access> <write-access> <namespace-resource root=""> <security-role-mapping name="<jndi-user-role>"> <group name="administrators" /> <group name="users" /> </security-role-mapping> </namespace-resource> </write-access> </namespace-access> ... </orion-application>
With this configuration, you can access remote EJBs without specifying principals or credentials.
To enable fat client access to EJBs using RMI, you must grant RMI permission "login" to the appropriate role. You can accomplish this in one of the following ways:
For the file-based provider, through Application Server Control by selecting the role and checking the "Grant RMI Permission" checkbox. (Also refer to "Create a Role" or "Edit a Role".)
Or through the OracleAS JAAS Provider Admintool:
% java -jar jazn.jar -grantperm myrealm -role myrole \ com.evermind.server.rmi.RMIPermission login
Restart OC4J for changes to take effect.
For an application to access EJBs, you must create the end user JDOE_ENDUSER
in the system-jazn-data.xml
file and grant it RMI permission "login".
This section discusses subject propagation in OC4J, and documents how to enable it with ORMI. (It is always used with IIOP, in accordance with the CSIv2 specification.) The following topics are covered:
Note: Subject propagation is a powerful feature that should be used only in environments where the server is secure from untrusted client access. It is therefore advised, in order to ensure proper integrity of client requests, that appropriate safeguards be established before this feature is used in production environments. For example, consider using application or network firewalls, RMI access restrictions (through the<access-mask> element in rmi.xml , as documented in "Configuring ORMIS Access Restrictions"), or RMI subject-propagation restrictions (through the <subject-propagation-mask> element in rmi.xml , as documented in "Removing and Configuring Subject Propagation Restrictions").
|
OC4J supports subject propagation, as summarized in Figure 14-4. Through this feature, a Web client can establish its identity to a servlet, and the servlet can then use that identity to communicate with other EJBs and servlets, where the identity is the appropriate JAAS subject (javax.security.auth.Subject
instance).
After the client's current subject is obtained, through a Subject.getSubject()
call, subject propagation works as follows:
The subject is serialized over to the RMI server.
The RMI server deserializes the subject and uses it to set up the server-side JAAS context.
Subjects may be propagated through a series of EJB invocations, for example. The EJB incorporates the client identity if the EJB is configured to use the client's identity. The EJB cannot be configured to use run-as mode with a specific role.
The authenticated subject (from supplied JNDI credentials) is then merged with the propagated subject to form a combined subject.
In OC4J, you can use subject propagation with ORMI if you specifically enable it on the client and server. You can accomplish this with the following system property setting at each end:
-Dsubject.propagation=true
In the current release, this setting controls subject propagation at a global OC4J level.
Be aware that for subject propagation to work properly, JAAS mode must be enabled for the Web application where the subject is being propagated from, and for the EJB where the subject is being propagated to. So for each, there must be a setting of jaas-mode="doAs"
or jaas-mode="doAsPrivileged"
in the orion-application.xml
file.
While java.security.Subject
is a class provided with the JDK, java.security.Principal
is an interface that you can implement as desired. For subject propagation to work properly with ORMI, you must ensure that the remote client, application, and OC4J all have access to any Principal
class definitions.
You can accomplish this by putting them in a library that is loaded as an OC4J shared library. There are two main steps to this (considering functionality of the Application Server Control Console in particular):
Load the library as an OC4J shared library. From the Application Server Control Console Administration tab for the OC4J instance, use the Shared Libraries task.
This results in configuration such as the following in the OC4J server.xml
file:
<application-server ... > ... <shared-library name="mylib.jar" version="1.0" library-compatible="true"> <code-source path="../mypath" /> </shared-library> ... </application-server>
Import the library into your application. In deploying an application through Application Server Control, when you reach the Deploy: Deployment Settings page (as discussed in "Deploying an Application through Application Server Control"), you have the opportunity to import shared libraries.
This results in configuration such as the following in your application orion-application.xml
file:
<orion-application ... > ... <imported-shared-libraries> <import-shared-library name="mylib.jar" /> ... </imported-shared-libraries> ... </orion-application>
Note: This is the preferred way to use shared libraries in OC4J; however, the<library> element and ORACLE_HOME /j2ee/home/applib directory are still supported.
|
See Also:
|
By default, access to subject propagation is denied to all ORMI clients. You can configure desired access through settings in the <subject-propagation-mask>
element and its <host-access>
and <ip-access>
subelements in rmi.xml
.
Subject propagation access can be either exclusive or inclusive:
In the exclusive mode, access is denied to all IP addresses or hosts except those specifically included. Use mode="deny"
in <subject-propagation-mask>
, then specify which particular hosts or IP addresses to allow by using mode="allow"
in a <host-access>
subelement, <ip-access>
subelement, or both.
In the inclusive mode, access is available to all IP addresses or hosts except those specifically excluded. Use mode="allow"
in <subject-propagation-mask>
, then specify which particular hosts or IP addresses to deny by using mode="deny"
in a <host-access>
subelement, <ip-access>
subelement, or both.
The following example configures an exclusive mode, allowing subject propagation for only localhost
and 192.168.1.0
. (255.255.255.0
is the applicable subnet mask.)
<rmi-server ... > ... <subject-propagation-mask default="deny"> <host-access domain="localhost" mode="allow"/> <ip-access ip="192.168.1.0" netmask="255.255.255.0" mode="allow"/> </subject-propagation-mask> ... </rmi-server>
The default setting is as follows:
<subject-propagation-mask default="deny"/>