A mini proof-of-concept OIDC federation using the Swamid/Amsterdam profile
This is an extremely simple federation with 2 entities. One RP and one OP. As configured both entities belong to the same federations. There are 2 federations/communities. The federation is SWAMID and the community is EduGAIN. SWAMID being a member of EduGAIN.
You should read the more detailed description that are part of the fedoidcmsg documentation. What I'll give you here will be a very short introduction.
The Swamid profile is built on the assumption that all entities in the federation are registered directly with the federation operator.
The Federation Operator (FO) is expected to run a metadata signing service (MDSS) that publishes the signed metadata statements for all entities within the federation.
The entities them self are not expected to keep signed metadata to be included in requests and/or responses.
You will need pip to get everything installed.
If you don’t have pip, this will help you with that: https://pip.pypa.io/en/stable/installing/
Using pip you must install:
- oidcop
- oidcrp
- fedoidcendpoint
- fedoidcrp
- atomicwrites
$ pip install X
should do it. Where X is one of the 4 packages listed above. !! You MUST get all of them !!
Once you have the 4 packages you can get the Swamid mini fed setup.
$ git clone https://github.com/rohe/oidc-swamid-federation.git $ cd oidc-swamid-federation
and read README.md But then you are already doing that so just keep on reading. The steps below must be done in order.
You have to create the Federation Operators keys. This you do by running create_fo_bundle.py . This script will create keys for the SWAMID and EduGAIN federations. Placing the private keys in private/ and the public keys in public/. It will also construct a FO bundle directory called fo_bundle . You MUST run the script from the root of the package.
The metadata signing service, run by SWAMID/SUNET, has to be initiated. There are more steps that involves the MDSS but they will be taken care off as they appear in the description below.
As mentioned above there are one federation and one federation community involved in this mini federation. They are each of the them described by their own configuration file (edugain_conf.json and swamid_conf.json).
Since the EduGain and Swamid keys where constructed by create_fo_bundle.py the configurations here only need to point out where the keys can be found.
The MDSS on the other hand needs it's own keys so the mdss_conf.json file will contain a specification of what type of signing keys the MDSS should have.
create_sign_seq.py is the script that will do the MDSS setup. Apart from constructing the MDSS keys it will construct a set of metadata signing sequences. One for MDSS/SWAMID and the other for MDSS/SWAMID/EduGAIN. One such set will be made for each of the different contexts where signed metadata statements can be used::
- Provider info discovery response (discovery)
- Client registration (registration)
- Client registration response (response)
Now that basic FO setup is done so let's bring in the entities.
To do enrollment each entity MUST construct a JSON document that looks like this:
{
"entity_id": "https://rp.example.com/metadata",
"signing_keys": <JWKS>,
"metadata_endpoint": "https://rp.example.com/metadata"
}
For an RP the entity_id and the metadata_endpoint should be the same.
To construct this information from the configuration specification you can use enrollment_setup.py which once run should have created the RP's signing keys and written a file called enrollment_info that contains the above mentions information.
The entity_id for an OP should be the same as the issuer ID and since there are demands on the format/issuerID from the OIDC specification. Using the issuerID as the metadata_endpoint is probably not feasible.
Because of the above if you run enrollment_setup.py in the OP directory the resulting enrollment_info will have the IssuerID as the entity_id but will have a metadata_endpoint that is not the same as the entity_id.
If we assume that the RP and the OP administrators send there enrollment application to the FO the next thing that happens will be on the FOs side.
I don't have a separate FO directory, separate from the MDSS directory. So everything the FO has will be in the same directory are the stuff needed by the MDSS.
Running enroll.py will basically copy the enrollment_info files from the entity (RP/OP) directories to a directory called entities. It will also copy the MDSS public signing keys to a file called mdss.jwks in the entities directories.
enroll.py RP OP
This process emulates the enrollment process that the FO has. In reality it will definitely be more complex. I leave the implementation of an enrollment process to the reader as an exercise.
Apart from wht enroll.py does one more thing has to be transmitted from the FO to the entity and that is information about the MDSS. In the configurations of the RP and the OP you can see these lines:
'mdss_endpoint': 'https://localhost:8089',
'mdss_owner': 'https://mdss.sunet.se',
'mdss_keys': 'mdss.jwks',
If you had read the SWAMID profile documentation in the fedoidcmsg documentation you would have seen that the endpoint the RP and the OP needs to know about is the endpoint where they can find information about the collection of signed metadata statements that the MDSS holds on the entity's behalf. The path is given in the specification to be:
/getsmscol/{context}/{entityID}
So only information about the leading part has to be transmitted. The RP/OP will also need to know about thepublic part of the signing keys the MDSS uses and the entity ID of the MDSS.
Now the FO should collect metadata from the newly enrolled entities. For this to work in this setting the RP and the OP needs to be running. You can start them like this:
$ cd RP
$ ./rp.py -t -k conf &
$ cd ../OP
$ ./server.py -t -k conf &
And then run the collector
$ cd MDSS
$ ./metadata_importer.py
After having run this you should have 2 files in a directory called entity_metadata. One called https%3A%2F%2F127.0.0.1%3A8100 and the other https%3A%2F%2Flocalhost%3A8080%2Fmetadata.
What metadata_importer.py did was look in the directory entities and for each file (entity) in that directory, based on the information in the file, it fetched the metadata from the metadata_endpoint (in the form of a signed JWT) verified the signature and stored the information in the JWT under the same file name as in the entities directory in the directory entity_metadata.
Now when we have the metadata for each entity there are a couple of things the FO has to decide on.
- Are there any claims that should be removed/modified
- Are there claims that should be added
- In which contexts should this metadata be used.
After all those things has ben considered/applied then a signed version of the metadata should be produced.
The three actions above have been modelled by the use of one directory with some rules process_rules and one script processor.py.
The files (one per entity) in the process_rules directory all follows the same pattern. And it looks like this:
{
"registration": {
"https://edugain.org": {
"federation_usage": "registration"},
"https://swamid.sunet.se": {
"federation_usage": "registration"}
}
}
The meaning of this is that this metadata can be used for registration in the EduGAIN and SWAMID federations. And one claim 'federation_usage' will be added to the metadata.
The processor.py will grab the metadata from the entity_metadata directory, apply the rules and create a number of files in the directory tree for processed metadata. The structure of this tree is:
{context}/{queue}/{federation_id}/{entity_id}
an example being
registration/in/https%3A%2F%2Fswamid.sunet.se/https%3A%2F%2Flocalhost%3A8080%2Fmetadata
There are 2 queues (in/out). processor.py will write in the in queue and the script signing_service.py will write to the out queue.
The next step in the chain is to construct signed metaddata statements for all the metadata instancecs that appear in the in queue. signing_service.py will do this for you. Just run it and it will populate the {context}/out directories.
Now we have all the necessary data bases in order so we should be able to run the MDSS.
$ ./mdss.py -t config
should do it.
If you've follow the instructions letter-by-letter you should have the RP, the OP and the MDSS running by now.
Point your Web browser at https://localhost:8080 and enter diana@localhost:8100 as you unique identifier.
Using diana/krall as username/password you should end up with a web page with some information about Diana. If not tell me what went wrong.