The basic requirement is to centralize the authentication and authorization of multiple SaaS applications to ease development (each SaaS application using minimal code to authenticate against a single source) and when necessary provide SSO. The authentication mechanism must handle the following options available to the user:
Use Third Party Authentication -- Google
Use our centralized authentication
Use the corporate provided authentication (ADFS)
In my research, I have found many, many ways this can be done and have found OpenAM to be the most complete solution, but then I came across FreeRadius which could also be used.
My Questions are:
There seems to be a plug-in for each tool where one can use the other together (OpenAM - authenticate against radius server), but is there any use case where FreeRadius would be preferred as the SOLE authentication server over OpenAM.
Does OpenAM require that a web agent installed for the server - if all I am doing is serving a Restful Interface (developed in Node.js) - is it possible to authenticate users without installing a web agent (there is no web agent for Node.js).
Can I pass user credentials from Browser -> Server (node.js) -> OpenAM thereby not giving the user the OpenAM login screen. The OpenAM token will be passed from OpenAM -> Server -> Browser (setting the cookies's origin as the SaaS's application.
That is each SaaS application server will serve as a "proxy" for user management (authenticate, authorize, and manage[create|update|delete] users)
Thank you
I'm early to the Open Identity Stack game but I am deploying an OpenAM (and OpenIDM + OpenDJ) based solution to handle exactly the solutions you mention.
direct answers:
As far as handing sole authentication over to FreeRadius I don't see why you would want to but anything is possible. Given your mention of the multiple directories (identity sources - google, ADFS, and your centralized authentication) I would think hooking up OpenAM to provide the RADIUS authentication (i.e. OpenAM RADIUS hook, not FreeRadius) would make sense.
No, a web agent doesn't have to be applied but it may make sense. There are some node.js pieces to help (https://github.com/alesium/node-openam). You just need to talk from your server to the OpenAM side (REST) and that should be good.
You can do that or you can just skin the OpenAM login screen to look like your own. I'd suggest the latter as you're then relying on OpenAM for the login screen security. If you're doing a pure proxy then you take that burden on. Your call as a design decision obviously.
good luck!
you're comparing a RADIUS sever with a Web SSO solution ... I'm not sure if this makes sense.
It seems FreeRadius does not have that many 'auth backends' (like Oauth to leverage Google Auth)
I am looking into the solution for a similar requirement myself, but I am looking to integrate 2FA as well. I have seen so many different solutions, but haven't pinned down the best one yet. Here is what I have come up with so far:
RCDev OpenID seems to be pretty comprehensive, and it is free for cases with less than 40 users.
Green Rocket's GreenRADIUS is expensive, but they have plugins for every scenario and it can work.
Red Hat's KeyCloak could be used in combination with TACACS+ or FreeRADIUS to accomplish this
Related
Getting started with Okta in a "security-conscious" organization, where connecting across the firewall requires a mound of paperwork.
Wondering if there are any best practices for testing applications internally before heading to the DMZ. There are two parts that we're primarily concerned with:
Authentication - any way we could emulate authenticating against something like an Okta preview env't without traversing the firewall?
API testing - while doing internal development, it would be great to be able to make API calls like "create user" against something that is purely internal - even if it's just a stub.
Okta doesn't currently offer a full-fledged mock Okta client, so there's no officially supported method for testing Okta locally.
There's a partial, undocumented mock-okta client here. However, it currently only supports the /oauth2/v1/authorize authorization code flow.
Can anyone Tell me if it is possible to combine SSO from Spnego and Spring security with Oauth
This is my problem :
The Client I now represent has chosen Spnego as their SSO solution.
This requires us to use a full blown appServer (Liberty) in all scenarios.
At the same time, the knowlegde and skills about Spnego in the developent team is very limited.
Due to issues with creating the keytab files, Spnego is only available in the formal test environment and not our local test enviroment.
This makes it very difficult/time consuming to test and devlop due to the long deployment time to the formal test enviroment.
Not over to my question:
If possible I would like to be able to "log in" to a service in the formal test enviroment (OAUTH2 authentication server ?) using SPNEGO SSO and get a token back that I can use in further requests towards my services located locally and/or in any other test enviroment.
Is this even possible ? I have not seen any examples where the authenticantionServer is using another sso provider to actually authenticate the user.
A different possibility might be to to do some sort of redirect from the login service in the test environment but I fear the Spnego token created only will be valid on a sever in the same domain..
I`m sorry if this question is confusing or not clear.
My knowledge of this domain (security) is limited and I struggle to get a grasp of how I can test my code locally with security enabled.
Links to any resources on the net that addresses some of these issues will be greatly appreciated.
I am to be having a lot of problems, misinformation and confusion when attempting to find out the plausibility and viability of attempting this.
The requirement is for a remote client, accessing our website to be auto signed in with their Active Directory User account.
We have the option to setup a WCF service (or something similar) on their remote server for authentication purposes. Which from my little understanding is how this problem will be tackled.
So, my question after a little background is this.
CAN this be done, and HOW can it be done?
Instead of hosting a WCF service on their domain, I would look into installing ADFS on their domain.
You can change your website to accept security tokens from ADFS using the WS-Federation protocol. You can use classes from the System.IdentityModel namespace for that. An example of how to implement this in ASP.NET can be found here.
An alternative would be to use Azure Active Directory as your identity provider and have your client sync accounts to their AAD directory (or federate between AAD and ADFS). An example can be found here.
Since Google lists the ClientLogin API as deprecated what is the recommended alternative?
In my case I need a server to authenticate and post things using credentials which I own. Alternatives like OAuth won't work in this context
The authentication method varies based on the scenario of your application. The scenarios reported by Google documentation are:
Login
Web Server Applications
Client-side Applications
Installed Applications Devices
Service Accounts
In your case, the Web Server Application scenario might be useful (if you give more details about what do you mean for "In my case I need a server to authenticate" I can be more precise about this point).
However, the recommended alternative to ClientLogin is OAuth2.
I am migrating one ASP.NET MVC 3 intranet Website to the Windows Azure and DB to SQL Azure.
On Premises my site uses Windows Authentication to authenticate and authorize the user(By Placing AUTHORIZE attribute on controllers).
It would be very kind of you If you can let me know How to go about the same.Thank You In Advance !
You have two choices here:
Use federated authentication and something like ACSv2. This requires a bit of work to setup a relying party, install ADFS2, etc. However, it is the most robust and future proof option. It is a very good option.
Use something like Windows Azure Connect. That will bring Windows Authentication to the cloud by joining your running instances to your domain controller on-premises. In effect, you have something of a VPN between your cloud instances and your on-premises domain controller. There are some caveats to this model today (requires installing agent on DC for instance), but it would be from a 'just works' stand point, the easiest. Longer term, this is less attractive I believe than option #1.
You can get more details for each of these by checking out the Windows Azure Platform Training Kit.
I should also add that you have no option (today at least) of using Windows Authentication with SQL Azure. You must use SQL authentication there, so what I am talking about here only applies to the web site itself.
I'm very successfully using Windows Identity Foundation with Azure AppFabric Access Control Service to authenticate using ADFS v2.
As well as straight authentication, it gives you lots of flexibility over other claims, such as roles (which don't need to be based solely on AD group membership).
In my opinion, its biggest strength is that there is no communication channel required between the Azure platform and your on-premise AD. Everything is done via the browser. From a security perspective, this means that although anyone can reach your application, nobody can authenticate to it unless they can also reach your ADFS server. Access to this can be restricted to on-premise clients only or via VPN, greatly reducing the attack surface.
Also, because ADFS does not need to be exposed externally, it can greatly ease the bureaucratic overhead of deploying it, in my experience.
Only configuration is required, which although it can be a bit of a fiddle initially, is pretty straightforward once you've got to grips with it. You configure WIF to use ACS as it's Identity Provider and create a Relying Party in ACS for the application. Then, you configure ACS to use ADFS as its Identity Provider. You could configure WIF to talk directly to ADFS, but the additional level of abstraction of going via ACS can be useful.
Once you've done your configuration, using the [Authorize] attribute 'just works'.
Note that if you're using Ajax calls into your controllers, you'll need to take some precautions, as Ajax calls don't handle the federated authentication redirect (or the ADFS Shuffle, as I like to call it), but it's nothing that's insurmountable.
All in all, I'm very impressed with the combination of WIF+ACS+ADFS for transparent Windows integrated authentication.