PingFederate redirect URL after logout

Readers

Logout is a typical requirement in Single Sign-On world. In the interest of today's post, PingFederate can perform SAML single logout or individual adapter logout and I will talk about these topics in detail in separate post.

It is also a common requirement to be able to redirect user to login page post logout. Today, I will explain the PingFederate capabilities to redirect user to different url post SAML logout.

PingFederate SAML logout end-point is https://<>/startSLO.ping, and this URL will perform SAML single logout. 

TargetResource parameter should be appended to logout URL to redirect the user to login page or to land onto different page.

https://<>/idp/startSLO.ping?TargetResource=https://<> 

This TargetResource should be whitelisted as shown below. 

Add the new <> under Security -> Redirect Validation section by specifying the domain name as abc.domain.com and select TargetResource for SLO and Other checkbox. This setting should be added only when ENABLE TARGETRESOURCE VALIDATION check box is enabled for SLO AND OTHER OPTION.


Without whitelisting the redirect URL, you will see below error 

Tomcat "Cannot recover key" SSL error

Tomcat has been enabled with SSL port. While renewing the SSL certificate, Tomcat server is throwing below error.

java.io.IOException: Cannot recover key
        at org.apache.tomcat.util.net.jsse.JSSESocketFactory.init(JSSESocketFactory.java:507)
        at org.apache.tomcat.util.net.jsse.JSSESocketFactory.createSocket(JSSESocketFactory.java:218)
        at org.apache.tomcat.util.net.JIoEndpoint.bind(JIoEndpoint.java:400)
        at org.apache.tomcat.util.net.AbstractEndpoint.init(AbstractEndpoint.java:650)
        at org.apache.coyote.AbstractProtocol.init(AbstractProtocol.java:434)
        at org.apache.coyote.http11.AbstractHttp11JsseProtocol.init(AbstractHttp11JsseProtocol.java:119)
        at org.apache.catalina.connector.Connector.initInternal(Connector.java:978)
        at org.apache.catalina.util.LifecycleBase.init(LifecycleBase.java:102)
        at org.apache.catalina.core.StandardService.initInternal(StandardService.java:559)
        at org.apache.catalina.util.LifecycleBase.init(LifecycleBase.java:102)
        at org.apache.catalina.core.StandardServer.initInternal(StandardServer.java:821)
        at org.apache.catalina.util.LifecycleBase.init(LifecycleBase.java:102)
        at org.apache.catalina.startup.Catalina.load(Catalina.java:642)
        at org.apache.catalina.startup.Catalina.load(Catalina.java:667)
        at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
        at sun.reflect.NativeMethodAccessorImpl.invoke(Unknown Source)

The error indicates that key can't be recovered using the password specified in server.xml. SSL section in server.xml is as shown below. If you have wallet pfx/p12 file, then you may convert it to java keystore using the same wallet password to use for java keystore password. 

<Connector port="8443" protocol="org.apache.coyote.http11.Http11Protocol"  maxThreads="100" SSLEnabled="true" scheme="https" secure="true"            keystoreFile="/u01/app.jks" keystorePass="XXXXX"  clientAuth="false" SSLProtocol="TLS" sslEnabledProtocols="TLSv1.2" />

The keystorePass value is different from passphrase used while creating private key/cert. After recreating the java keystore with same password as key and updated server.xml with new password and Tomcat started working fine. 

Azure AD SAML Integration issue with PingFederate : AADSTS75011

Readers, I'm back!

I would like to resume blogging my working experiences and best possible solutions on technologies that I come across every day.

Today, I would like to share some limitation that I came across on Azure AD SAML integration.

Scenario:

A cloud application is integrated with PingFederate where Ping is IDP and app is SP.
PingFederate is delegating authentication to Azure AD; so Ping is SP and Azure AD is IDP.

Issue:
As per the scenario, all three systems Application, PingFederate and Azure AD are configured with SAML integration.

When Application is accessed, it is throwing an error when it is landing on Azure AD for authentication.

Cause:
As per Azure error code AADSTS75011, the problem seems to be Authentication Context parameter sent by the Application in SAML Authentication request as shown below.

<saml:AuthnContextClassRef xmlns:saml="urn:oasis:names:tc:SAML:2.0:assertion">urn:oasis:names:tc:SAML:2.0:ac:classes:PasswordProtectedTransport</saml:AuthnContextClassRef>

Refer this article, https://techcommunity.microsoft.com/t5/azure-active-directory/aadsts75011-by-which-the-user-authenticated-with-the-service/m-p/196130#

Azure AD supports only urn:oasis:names:tc:SAML:2.0:ac:classes:Password as authentication context value in SAML requests. Refer this article, https://docs.microsoft.com/en-us/azure/active-directory/develop/single-sign-on-saml-protocol

Fix:
Ensure that application would send AuthnContextClassRef parameter as urn:oasis:names:tc:SAML:2.0:ac:classes:Password

Best practices for Oracle Access Manager deployment

Readers,

It is very important to follow best practices for OAM deployment for either single DC setup, or MDC setup or HA functionality etc., Some of the best practices are mentioned below and I will keep updating this post based on comments/feedback.

1. Verify the certification matrix for OAM deployment before beginning installation.
2. Use the latest Java, WebLogic Servers based on certification matrix.
3. Use the same path for Weblogic Home, Oracle Home and Oracle Common Home across environments.
4. For WebServers, use environment specific instance names rather than default names. For eg., while installing and configuring OHS instances, the default instance name is ohs1. It is suggested to use name denoting environment name and server name such as DEV_OHS1 or PROD_OHS1. 
5. For WebServers in HA, if you are installing OHS on multiple boxes then it is suggested to use different OHS instance names. For eg., PROD_OHS1 (on 1st box), PROD_OHS2 (on 2nd box). This will help associating OHS instances with unique instance names in EM console for monitoring purpose. 
6. Ensure that Non-Prod (or QA) and Prod environments identical for all IAM components in order to reproduce Production issues and make future deployments easier and better for operations.
7. Ensure to configure OMSM, Policy Manager and OAM servers in different clusters during initial WebLogic domain configuration.
8. Use load balancers for backend LDAP servers for user store. 
9. Use custom authentication plugins only as per needs. Custom Plugins have maintenance nightmares for Ops team. 
10. Schedule regular backups for oam-config.xml, policy exports and partners export along with DB backups.
11. While configuring user store, specify Referral policy as FOLLOW. Also, specify Min/Max pool size and timeout parameters as per load testing results. 
12. Selectively configure required audit events only.  
13. Avoid creating sample cgi or server side pages that would print headers information. For eg., cgi-bin/printenv script is a common script that's protected through OAM for validation purpose. After Production go-live, remove those test pages. 
14. Exclude images, css files and other static contents in policies instead of using Unprotected policies. 
15. Try to update WebLogic CPU patches and JDK CPU patches based on sensitivity of the fixes. Thorough testing is required before pushing these patches to Production. Sometimes these patches may break OAM functionality. 
16. Use similar RCU versions across environments.
17. Use certain security headers X-XSS-Protection, Strict-Transport-Security etc., in Login WebGate WebServers for better protection. 
18. Use Log rotation for OAM and other WebLogic servers to prevent complete disk usage. 
19. Ensure that only required firewall ports are opened to and from OAM servers instead of opening it through entire subnets. This is to provide better security.
20. Use WebGate CERT mode only if needed otherwise it will degrade performance. 
21. Use SSL at Node Manager and WebLogic Servers are only as needed. 
22. Ensure that Login WebGate has SSL enabled.
23. Ensure that 3rd party cert is used for OAM deployment in Prod and non-prod environments. 
24. Do periodic rolling restarts of OAM server for better system performance.
25. Associate OHS and OAM components with EM 12c for performance and system alerts.
26. Use Min / Max Heap size for OAM tuning. 
27. Use appropriate number of processes, sessions etc., parameters in OAM DB schema.
28. If there are more IAM components in the architecture, it is suggested to use separate DB for OAM component since OAM uses APS for MDC synchronization where as other products requires DG or GG.
29. Ensure to grant admin privileges to LDAP users for OAM console instead of giving generic administrative access to all users.  

 That's it for today. Happy reading!! 

Enabling Multi factor Options for application in Okta

Okta is a simple product to use and it is capable of integrating with more than 5800 applications for Single Sign-On purposes with added security benefits.

In this post I would like to quickly talk about enabling Multi Factor capabilities for an application that is already integrated into Okta. In this case, SalesForce is integrated with Okta using SAML 2.0 protocol.

Follow the below steps to enable MFA capabilities for salesforce application (or this would apply for any other application)

1. Login to Okta Admin Console as admin user.
2. Goto Applications.
3. Goto Sign On Policy section.
   
4. Click Add Rule.
5. Specify Rule Name, Conditions (users /group rule assignment), Location . 
   
6. Specify Actions. Select the option to allow or deny the access after conditions are met. For this exercise I would choose Allow. Select MFA frequency such as prompting MFA every login or once a day etc.,
7. Click on Multifactor Settings.
   
8. Select the desired authenticators. 
   
9. Click Save. 

Now the desired configuration changes are completed. It is time to test. 

1. Login to Okta application console. 
2. Click on SalesForce application.
   
3. User is redirected to MFA page for setup (if its already setup, this page is not seen). In this case, I have selected Google Authenticator setup. Click Setup.
   
4. Select device Type. Install app on mobile device. Click Next.
   
5. Specify the code generated on mobile device here. Click Verify.
   
6. User is redirected to salesforce application. 
   

PasteConfig failed in OAM MDC

Do you have OAM, OMSM and Policy Manager servers in your environment? Are they all tied to same OAM WebLogic cluster or different clusters?

Well, the recommended approach is to tie respective weblogic server to respective cluster. Therefore, OAM servers should be associated with OAM Cluster, OMSM servers should be with OMSM cluster and so are Policy manager servers.

This post will explain about an issue that you may come across when WLS servers are associated in their respective cluster.

I will skip the steps for T2P commands on the Master DC as they are not relevant to this post. After running extraceMovePlan.sh script on Clone DC, it generates moveplan.xml which is the basis for pushing configurations into Clone DC and creating weblogic domain.

moveplan.xml contains server details and cluster details etc., for OAM servers, OMSM servers and Policy Manager servers.

The default xml template would look like as shown below.

With this moveplan.xml, pasteConfig is failed.

After commenting out the omsm and policy manager server definitions in moveplan.xml, pasteConfig has worked.

Also, the weblogic domain was created for omsm and policy manager servers even though it was commented in moveplan.xml.

Specifics of Multi Data Center using Oracle Access Manager

Readers,

This is a very generic post for understanding some specifications of designing and implementing Multi Data Center solution using Oracle Access Manager.

• If you are designing an IAM solution in a single Production data center involving multiple IAM products, then it is mandated for separate Database for OAM component alone. This is because OAM uses Automated Policy Synchronization feature to sync artifacts/policies from Master to Clone DCs without the need of database replication unlike OIM and OAAM etc., Therefore, if OAM and other products are using same database then OAM database should be separated before deploying a new OAM data center. 
• In a typical IAM deployment, login pages are either served from OAM or through DCC. In both case, front end web server is employed with WebGate. It is preferred to use Global LBR for OHS data center along with load VIP. Therefore, a global LBR will also be assigned as OAM server hostname. If N number of data centers are added in the future, a new local LBR can be assigned to Global LBR without changing the OAM hostname.
• It is advised to employ Global LBR for OAM servers (default 5575 TCP port) for WebGates to OAM server connections. 
• Server affinity and sticky session is mandatory for all Global LBRs in a MDC solution. 
• OAM to database connection should have direct connection instead of Global LBR for all databases since each OAM talks to its respective DC database.
• In the event of using Virtual Host concepts at web server layers, it is advised to have Global LBR with sticky session and server affinity for OAM server 14100 port too. 
• OAM LBR for 5575 port should be TCP and OAM LBR for 14100 port should be HTTP.
• If mod weblogic plugin is used in the deployment for MDC, it is advised to use OAM LBR for 14100 port using WebLogicCluster parameter. Some of the nuances of using WebLogicHost and WebLogicCluster for mod weblogic plugin will be explained in detail in separate post.
• Before running T2P commands on Master DC, it is mandated to remove IAMSuite agent password and mark Embedded LDAP as system /default stores.
• In an MDC solution, Primary DC OAM is designated as Master thus allowing write operations only in this OAM server. All the policy updates from Master DC will be pulled into Clone DCs using APS feature.
• When a new DC to be brought into the existing MDC deployment, perform T2P on master DC to generate binaries rather than using old T2P binaries because something might have changed since initial T2P deployment. 
• Ensure to run pasteConfig and pasteBinary only on Administrator node of OAM component and use pack/unpack for other OAM nodes.
• It is usually recommended to have single Middleware on a host to avoid jar conflicts etc., even though we had seen OAM and OUD binaries co-existing on same Middleware. 
• It is always advised to add a new OAM server on separate host instead of adding OAM server on an existing OAM host as this would provide better CPU and server performance. However, this can be decided after thoroughly analyzing performance metrics. 
• Attain due diligence with Transformation rules as incorrect rules or incomplete rules will result in abnormal behaviors and corrupt the Clone DC environment in the worst scenarios. Thus, thorough testing with APS is required.
• Session control parameters define the user behavior during run time in an MDC deployment. Therefore, thoroughly test the MDC solution using these parameters. Below parameter values allow users to access SSO protected applications even though the sessions are created in Remote OAM DCs by creating local sessions in the event of session retrieval failures. 

          SessionMustBeAnchoredToDataCenterServicingUser=true

          SessionDataRetrievalOnDemand=false

          Reauthenticate=false

          SessionContinuationOnSyncFailure=false

• Designing Global LBRs at every component OHS, OAM and LDAPs will provide 100% high availability and failover even when one of the component in a data center fails. 
• Ensure that firewall connectivity is open from one DC component layer to all other DCs as cross over functionality.
• Perform the load testing in MDC deployment before going live as this will provide the SLAs and response time for each and ever component and its LBR to fail over to other available local VIP.  
• There is no MDC support for OAM OAuth tokens. Therefore oauth access token generated in one data center is not validated by remote DC. 

As always, TEST, TEST , TEST the component MDC solution.