ZXID.org Identity Management toolkit implements standalone SAML 2.0 and Liberty ID-WSF 2.0 stacks. The IdP provides Single Sign-On capability and the Discovery provides ability to issue credentials to web services as well as to locate suitable services.
ZXID IdP and Discovery configuration and databases are filesystem based and consists mainly of text files to which lines are appended. After describing the filesystem based configuration, we introduce some IdP administration topics and how to accomplish them.
mod_auth_saml Apache module documentation: SSO without programming.
zxid_simple() Easy API for SAML
ZXID ID-WSF API: Make Identity Web Services Calls using ID-WSF
ZXID Configuration Reference: Nitty gritty on all options.
ZXID Circle of Trust Reference: How to set up the Circle of Trust, i.e. the partners your web site works with.
ZXID Logging Reference: ZXID digitally signed logging facility
javazxid: Using ZXID from Java
Net::SAML: Using ZXID from Perl
php_zxid: Using ZXID from PHP
zxididp: Using ZXID IdP and Discovery
README.smime: Crypto and Cert Tutorial
FAQ: Frequently Asked Questions
README.zxid: ZXID project overview
zxididp implements, in a bare-bones way, SAML 2.0 IdP, ID-WSF 2.0 Authentication Service, and ID-WSF 2.0 Discovery Service. It is one of the cornerstones of a federated identity architecture.
First you need the following binaries
You can get all of these from the ZXID package by compiling it. Just say `make'
You also need a web server that can run CGI scripts. For simplicity, I recommend mini_httpd.
You need to decide what URL the IdP will live on. Typically the URL will end in zxididp, but you can change this by renaming the zxididp binary. You need to register the domain name and obtain certificate for it, unless you are willing to use the self signed certificates that ZXID generates. This tutorial assumes you chose following URL
N.B. Prior to version 1.10 (early 2012) zxididp would use /var/zxid/idp as default PATH and /var/zxid/idpzxid/conf as default configuration file. Since implementation of VPATH and VURL this is no longer supported. We recommend either using virtual hosting - with PATH ending up something like /var/zxid/idp.example.com:8443/ and configuration file in /var/zxid/idp.example.com:8443/zxid.conf - or deploying the IdP as sole application on its server and using nonvirtual PATH=/var/zxid/ and configuration file at /var/zxid/zxid.conf. The rest of this chapter assumes the latter approach.
sudo ./zxmkdirs.sh /var/zxid/ sudoedit /var/zxid/zxid.conf NICE_NAME=Your IdP Name for End Users to See ORG_NAME=Your org ORG_URL=http://example.com/ BUTTON_URL=/example-saml2_icon_150x60.png URL=https://idp.example.com/zxididp IDP_ENA=1 AS_ENA=1
Assuming you have obtained a certificate, here's how you would install it. We assume you have the certificate in file cert.pem and the private key in priv.pem.
sudo su cat cert.pem priv.pem >/var/zxid/pem/ssl-nopw-cert.pem # put both in one file cp /var/zxid/pem/ssl-nopw-cert.pem /var/zxid/pem/sign-nopw-cert.pem cp /var/zxid/pem/ssl-nopw-cert.pem /var/zxid/pem/enc-nopw-cert.pem cp /var/zxid/pem/ssl-nopw-cert.pem /var/zxid/pem/logenc-nopw-cert.pem cp /var/zxid/pem/ssl-nopw-cert.pem /var/zxid/pem/logsign-nopw-cert.pem chmod 600 /var/zxid/pem/* # end su
cd zxid-1.10 sudo mkdir /var/zxid/webroot sudo mv *.html *.css /var/zxid/webroot sudo cp zxididp /var/zxid/webroot
Metadata exchange introduces SAML SP to SAML IdP. Among other things, the certificates are exchaged and trust relationship is formed. ZXID supports fully automatic metadata exchage ("Auto-Cot"), see MD_FETCH=1 config option. Here we show how to exchange metadata manually.
Generate our own metadata for exporting to SP
sudo zxcot -ci -m >idp-meta.xml
Alternately the SP can access the metadata at URL https://idp.your-org.com/zxididp?o=B
For each SP metadata needs to be obtained and imported. If the SP supports well known location method, then accessing SP's entity ID as a URL will return the metadata. ZXID based SPs support this method. Typically the URL ends in "?o=B". If SP metadata is available from a URL, you would import it like this:
zxcot -ci -g https://sp.example.com/saml?o=B
Alternately you can obtain the metadata as a file and transfer it to the IdP machine. If SP is ZXID based, you can create its metadata with command
zxcot -m >sp-meta.xml
Then to import it to IdP, run
sudo /usr/local/bin/zxcot -ci -a <sp-meta.xml
head /dev/urandom | sudo /usr/local/bin/zxpasswd -n .all # dummy user to represent commonality sudo /usr/local/bin/zxpasswd -at 'cn: Nomen Nescitur$email: email@example.com$o: Your Org' -n nn # Password will be asked on console
zxcot -e https://idp.example.com/zxididp?o=S 'TAS3 Default Discovery Service (ID-WSF 2.0)' https://idp.example.com/zxididp?o=B urn:liberty:disco:2006-08 | zxcot -bs
zxcot -e https://sp.example.com/saml?o=S 'Your service human readable description' https://sp.example.com/saml?o=B urn:yoursvc:2011-08 | zxcot -b
cd /var/zxid/webroot mini_httpd -u root -p 443 -c 'zxid*' -S -E /var/zxid/idppem/enc-nopw-cert.pem
By design the idp will not start automatically after reboot. Operator intervention is required to supply the password for the encrypted filesystem.
sudoedit /etc/idp-start #!/bin/sh # 20110618, firstname.lastname@example.org modprobe fuse /usr/local/bin/encfs /var/.zxid /var/zxid # you will need to supply password manually cd /var/zxid/webroot /usr/local/bin/mini_httpd -u root -p 443 -c 'zxid*' -S -E /var/zxid/idppem/enc-nopw-cert.pem -l /var/zxid/log/mini_443.log & #EOF sudo chmod a+x /etc/idp-start
Then just run
sudo /etc/idp-start # you will be asked for encfs password, which you must know
zxididp uses local idpuid/ directory to implement the IdP logins and to store users federations, bootstraps, and attributes.
/var/zxid/ | +-- idpzxid.conf Main configuration file +-- idppem/ Our certificates +-- idpcot/ Metadata of CoT partners (metadata cache) +-- idpses/ Sessions | | | +-- SESID/ Each session has its own directory | | | +-- .ses The session file | `-- SVC,SHA1 (Each bootstrap is kept in its own file) | +-- idpuid/ Local user ID (local login name) to SHA1 mapping | | | +-- JOE/ Local user has directory whose name is the login uid | | | | | +-- .log Log of operations regarding the user | | +-- .pw User's local password | | +-- .yk User's yubikey shared aes128 secret in hex | | +-- .ykspent/ Cache of already used One Time Passwords | | | | | | | `-- OTP File name is the spent Yubikey ticket (w/o uid) | | | | | +-- .bs/ Directory of bootstraps to be included | | | | | | | +-- .at Attributes to be included in each SSO | | | `-- SVC,SHA1 Bootstrap for a service | | | | | `-- SP,SHA1/ One directory for each SP user is federated with | | | | | +-- .mni Federated name id to be used with this SP | | +-- .at Attributes to be included for this SP | | `-- SVC,SHA1 Bootstrap to be included for this SP | | | `-- .all/ Template used for all users | | | +-- .bs/ Directory of default bootstraps to be included | | | | | +-- .at Attributes to be included in each SSO | | `-- SVC,SHA1 Bootstrap for a service | | | `-- SP,SHA1/ One directory for each SP | | | +-- .at Attributes to be included for this SP | `-- SVC,SHA1 Bootstrap to be included for this SP | +-- idpnid/ Index of federated NameIDs, to map to uid | | | `-- SVC,SHA1 Bootstrap to be included for this SP | | | `-- NID Content of the file is uid | +-- idpdimd/ Discovery Metadata registrations | | | `-- SVC,SHA1 Discovery MD registration for a service (an EPR) | `-- idplog/ Log files, pid files, and the like
zxpasswd(8) is a user provisioning tool that allows creation of new accounts as well as manipulation of .pw and .yk files.
When generating a SSO assertion, the attributes are collected as follows:
LDIF at /var/zxid/idpuid/JOE/.bs/.at
LDIF at /var/zxid/idpuid/JOE/SP,SHA/.at
LDIF at /var/zxid/idpuid/.all/.bs/.at
LDIF at /var/zxid/idpuid/.all/SP,SHA/.at
As of version 0.33 (20090904) the attributes are rendered singlevalued. If multiple occurrances of an attribute happen, the first instance is used and others ignored. However, in a future version, we expect to support multivalued attributes.
Keep attributes, that are not shared by default, in .optat file.
Support personas using files like .at-persona and .optat-persona where persona is an alias that user has chosen to assign.
Choice of persona is remembered at session level in file /var/zxid/idpses/.persona. The attributes of .at are augmented and overridden by the attributes in .at-persona file. .at-persona can also deny release of an attribute that would have been released by virtue of .at.
For each SP, the user can specify .at, .optat, .at-persona, and .optat-persona files. The SP specific files can deny attributes that would otherwise have been released by virtue of global .at and .at-persona files.
Finally, the user can choose to suppress or override the value of an attribute just for a specific transaction. The overriding value is not saved to SP specific, persona, or global layer unless user expressly requests so.
For each attribute the authority is determined. Default is self asserted. If attribute has other authority, this will be indicated. If user chooses to override the attribute, either for transaction or for a persona, its authority reverts to self asserted. Attributes from a given authority are provided in files of type .at,authority or .optat,authority.
Some attributes are reserved: user does not have the option of setting their value (i.e. self asserted is not an option), and depending on the case the user may or may not be able to suppress their supply. The unsuppressible attributes are listed in configuration option MANDATORY_ATTR, which is global (not per user or not per SP).
Attribute Release Algorithm
Initialize attribute set from internal defaults
Attribute Selection screen is displayed whenever
User has checked the Adjust attribute sharing -checkbox prior to authentication.
SP requests more attributes than default or SP specific attribute sharing policy permits. If SP does not specify which attributes it wants, then SP specific and default attribute sharing policy determine the attributes to release.
User has requested as a global preference (and this preference has not been overridden in a session) that all attribute releases be displayed.
User has requested at session level that all attribute releases be displayed.
User has requested for the SP that all attribute releases be displayed (similar to specifying an empty, but defined release policy towards SP, see point (b); difference is that the sharing defaults still tick the share boxes, providing convenience to the user.
Attribute release at discovery or web service call time is governed first by SP specific attribute sharing policy and then by the default sharing policy. If WSP needs more attributes, it will have to submit a web service request for them. Eventually, we may support using interaction service to query user about release of the extra attributes.
The bootstrap attributes (ID-WSF 2.0 EPRs with credentials) are attached to the SSO assertions to facilitate frontend web sites making backend web services calls.
The bootstraps are generated by scanning /var/zxid/idpuid/JOE/.bs directory and for each corresponding file looking up the service EPR in the /var/zxid/idpdimd directory. If metadata for the service is found there, an access credential for accessing the service is generated (unless disallowed by the configuration). The access credential is based on existing federation, or if that is lacking and auto federation is allowed, on newly generated (IdP unilateral) federation.
Future versions (as of 2009) may allow more granular approach to the bootstraps generated, but the strategic direction is to technically enable communications automatically and then to rely on explicit authorization services (XACML PDP) to make a policy based decision about what is allowed to happen.
After user's .bs directory, the global /var/zxid/idpuid/.all/.bs is scanned. This allows common bootstraps, such as discovery, that occur for all users to be specified only once at the .all level, while additional bootstraps can be added for the individual users.
N.B. As of 0.41 (20091120), the .di directories are unused. In case of discovery, all matching entries in /var/zxid/idpdimd will be considered and federations automatically created, irrespective of user's .di or .all/.di.
Yubikey support works by using the initial part of the ticket (passed in as user field) as uid and the latter as the ticket proper. The uid part is used to locate correct directory. Mapping from yubikey modhex to real UID is done by creating a symlink. The AES128 shared (between yubikey token and IdP) key is kept in the .yk file. As this is not a password has, but rather directly the shared secret, it requires rigorous approach to the filesystem security. The fact that .pw and .yk are separate files caters for the possibility of user authenticating either by yubikey or by password. By default yubikey is one factor authentication (in fairly secure and very convenient form). If two factor authentication is desired, the password component should be prefixed to the UID component, i.e. first user types PIN and then presses yubikey to add UID and ticket.
To program yubikeys with the shared secrets, you need ykpersonalize(8) tool, available from Yubico as open source.
TLS Client Certificate authentication of users has not been implemented yet, but in any case would be mainly implemented by configuration of web server to request such certificate and verify it. By the time zxid gets called, the client cert authentication will already have happened. HTTP Basic authentication works in similar way and we make no attempt to cater for it, although it can be used of configured separately (in the traditional way).
First you need to obtain zxididp either by downloading a binary package (and installing it) or by compiling from source. See zxid-install.pd for further details.
zxididp(8) is a CGI script (written in C) that you can install under any web server that supports CGI scripts (e.g. Apache httpd, IIS, or mini_httpd). You must configure your web server to execute zxididp(8) as a CGI script. The specifics vary from web server to web server - consult your server documentation.
For Apache this would mean in httpd.conf a stanza like
<Location "/zxididp"> Options All SetHandler cgi-script </Location>
You would then copy the zxididp binary to the root of your document tree and make sure it is executable. See apache.pd for full scoop.
For mini_httpd this could mean something like
mini_httpd -p 8443 -c zxididp -S -E zxid.pem
For IIS this could mean something like
In Microsoft IIS 7 "Handler Mappings" use "Add Script Map" and associate Request Path: "zxididp" to
C:/var/zxid/bin/zxididp.exe # Or what your path is
This is saved in file web.config
In IIS 6 add mapping for zxididp to run as CGI. Go to webroot -> Properties -> Virtual Directory -> Configuration -> Application Configuration
zxididp C:\var\zxid\bin\zxididp GET,...
Remember to add zxididp to Web Services Extensions New Web Service Extension, Add Required files
In IIS3 and earlier it was necessary
regedt32.exe HKEY_LOCAL_MACHINE/SYSTEM/CurrentControlSet/Services/W3SVC/Parameters/ScriptMap Create new value name "zxididp" and value "c:/var/zxid/bin/zxididp.exe"
You must choose location for the configuration and logging hierachy of the zxididp(8) and then populate it. See earlier section in this document for details. By default the hierarchy will live in /var/zxid and the directories will have prefix "idp".
/var/zxid/ | +-- idpzxid.conf Main configuration file +-- idppem/ Our certificates +-- idpcot/ Metadata of CoT partners (metadata cache) +-- idpses/ Sessions +-- idpuid/ Local user ID (local login name) to SHA1 mapping +-- idpnid/ Index of federated NameIDs, to map to uid +-- idpdimd/ Discovery Metadata registrations +-- idpinv/ Invitations `-- idplog/ Log files, pid files, and the lik
You can create this hierarchy by running
make dir ZXID_PATH=/var/zxid/idp
See zxid-cot.pd for full scoop.
If Auto-CoT is enabled, and SP and IdP can communicate over the internet (n.b. problems with DNS, firewalls, other party not supporting WKL, etc.), the Service Providers can join the CoT automatically upon first use.
Otherwise, metadata exchange may need to be arranged manually using zxcot(8) tool. Typically you would receive sp-metadata.xml by mail from the SP administrator, or by instant messaging file transfer. You could even fetch it yourself using a web browser, saving the metadata page. Metadata is a XML blob.
zxcot -a /var/zxid/idpcot <sp-metadata.xml zxcot -g https://site.com/meta.xml /var/zxid/idpcot # Assumes connectivity zxcot /var/zxid/idpcot # Lists the SPs in the CoT
When all else fails, you can just manipulate the filesystem manually, creating the necessary files and directories.
If SP and IdP can communicate over the internet (n.b. problems with DNS, firewalls, other party not supporting WKL, etc.), the Service Providers can just request the zxididp metadata on the fly, based on Well Known Location (WKL).
The EntityID and WKL of the zxididp(8) are of the form:
where the "o=B" part is the ZXID general way of triggering metadata response.
If you need to perform the metadata exchange manually, your best bet is to download the metadata with your web browser (see WKL above) and send it to the other party.
One way to facilitate adding new SPs to the Circle of Trust is to deploy the zxidcot.pl Web GUI. This allows the administrators of SPs and WSPs to add their metadata and discovery registrations as well as view the existing registrations. This self help method is suitable if you want to run very open CoT, i.e. anyone can join. It may not be appropriate if you want to control who enters your CoT.
To install zxidcot.pl, just copy it to suitable location in your web root and make sure it can be executed as a CGI script. If you keep your zxididp configuration in any place other than the default /var/zxid/idp, you will need to edit zxidcot.pl to correspond to your configuration. The visual aspects of zxidcot.pl can be controlled to some extent by editing the an.css cascading stylesheet, but if you need big changes, you will need to edit the script itself.
To access zxidcot.pl point your web browser to https://idp.example.com/zxidcot.pl?op=md for metadata registration or https://idp.example.com/zxidcot.pl?op=direg for discovery registration.
The underlying implementation uses the zxcot command line tool.
User provisioning and administration: Currently all this is manual and commandline or filesystem based. Contribution in form of fancy user self provisioning web GUIs would be appreciated by the zxid.org project.
Generally you should be able to administer the users using zxpasswd(8), but when all else fails, you can just manipulate the filesystem manually, creating the necessary files and directories.
Adding a new user consist of creating a directory corresponding to the user ID under /var/zxid/idpuid hierarchy. Typically this directory will contain the .pw file for password authentication and possibly a .yk file for Yubikey authentication. The directory will also contain .bs/.at file for specifying user's attributes.
Easiest way to create a new user is using -n option:
zxpasswd -n user /var/zxid/idpuid <passwd
zxpasswd user /var/zxid/idpuid <newpasswd
N.B. By default the password is stored as MD5 hash. Use -h X command line option to choose other hashing algorithm.
Yubikeys are One Time Password (OTP) hardware tokens that connect to the USB port and act as keyboard emulators. They are sold by yubico.com
zxpasswd -h y -s user yubikeyalias /var/zxid/idpuid <yk_aes128_secret
N.B. To program a Yubikey and agree to a AES 128 bit shared secret, you need ykpersonalize program, available from Yubico.
There is no command line interface for adding the attributes currently (20091120). You simply have to edit the UID/.bs/.at, UID/SP,SHA1/.at,
.all/.bs/.at, or .all/SP,SHA1/.at files. The files are in (relaxed) LDIF
dn: cn=Koerkki,o=Labra permisRole: teacher cn: Koerkki o: Labra
N.B.: The LDIF format is essentially an attribute name separated from an attribute value by colon and one space, and name-value pairs separated from each other by a newline. This is very similar to mail header format. The "dn:" line is required in LDIF, but can be omitted in zxididp .at files.
If you want to add a complex binary or XML attribute, such as a x509v3 attribute certificate or a SAML assertion, we recommend that you safebase64 [RFC3548] encode the attribute value to prevent it from being mangled by SP SSO processing. To unravel such safebase64 encoding you can use in mod_auth_saml, or other zxid based SP, the INMAP specification with rule as unsb64-inf or unsb64. See zxid-conf.pd, section "INMAP specification", for further information.
The attributes from all the available sources are accumulated.
If you are willing to accept new users on self asserted basis, you can set up zxidnewuser.pl script so that anonymous users can register themselves. If you do not want anonymous users, you should not be setting this script up. To control the visual appearance of the new user registration, you can edit an.css stylesheet or newuser-main.html template. To control which attributes are collected, you need to edit both the template and the zxidnewuser.pl script.
Fig-1: New user registration screen for self asserted users.
To install zxidnewuser.pl, simply place it somewhere in the web root of your website and make it executable as a CGI script. Then you need to add to /var/zxid/idpzxid.conf
You may also want to edit the an-main.html template to control how the new user button will appear.
If new user registration is called in middle of a SSO flow, zxidnewuser.pl will preserve the ar= CGI variable (as a hidden field) so that when new user has been created, the flow returns to the login screen and the new user can login. zxidnewuser.pl can also be invoked outside login sequence: just provide a link to it from the first page of the IdP. In this case after the registration, the user is returned to the top level index page of the web site.
The underlying implementation uses the zxpasswd command line tool.
N.B. Before this step you should check that the SAML metadata of the WSP is registered at the IdP, see zxcot -a, etc. Or make sure the auto cot feature is enabled.
The principal step in configuring bootstraps and discovery is to register service's EPR in /var/zxid/idpdimd directory. This can usually be done on command line as follows:
zxcot -e http://sp.tas3.pt/mysvc?o=S 'My Service' \ http://sp.tas3.pt/mysvc?o=B urn:x-mysvc \ | zxcot -b /var/zxid/idpdimd
Or if we analyze by parts
zxcot -e SOAP-Endpoint Description EntityID ServiceType > epr.xml
This invocation is a mere utility for generating an EPR, without credential, that can be used as the registration. This might look like
<wsa:EndpointReference notOnOrAfter="2037-01-05T23:03:59.001Z" wsu:Id="EPRID92lFPo3ZNEt_3rHtJFoU" xmlns:wsa="http://www.w3.org/2005/08/addressing" xmlns:wsu="http://docs.oasis-open.org/wss/2004/01/oasis-200401-wss-wssecurity-utility-1.0.xsd"> <wsa:Address>http://sp.tas3.pt/mysvc?o=S</wsa:Address> <wsa:Metadata> <di:Abstract xmlns:di="urn:liberty:disco:2006-08">My Service</di:Abstract> <di:ProviderID xmlns:di="urn:liberty:disco:2006-08">http://sp.tas3.pt/mysvc?o=B</di:ProviderID> <di:ServiceType xmlns:di="urn:liberty:disco:2006-08">urn:x-mysvc</di:ServiceType> <sbf:Framework version="2.0" xmlns:sbf="urn:liberty:sb" /> </wsa:Metadata> </wsa:EndpointReference>
The next step is to provide this as "discovery metadata":
zxcot -b /var/zxid/idpdimd < epr.xml
which will create an entry in the /var/zxid/idpdimd directory (n.b. specifying the directory explicitly is necessary because the default directory corresponds to the SP directory layout, now we want the IdP directory).
Once the service's EPR (Metadata) exists in the system, any user can discover the service, with federations being created automatically (unless disabled in the configuration, see MD_FETCH and MD_POPULATE_CACHE config options).
If you want the service to be included as a bootstrap, you need to mention it either in global /var/zxid/idpuid/.all/.bs directory or in user's .bs directory. The file in .bs directory has to have the same name as in /var/zxid/idpdimd directory. Contents of the file do not matter. You can either touch(1) the file in existence or supply -bs flag to zxcot(1):
zxcot -bs /var/zxid/idpdimd < epr.xml
This registers the bootstrap under .all user.
When all else fails, you can just manipulate the filesystem manually, creating the necessary files and directories.
By far the most common type of bootstrap is a discovery bootstrap:
zxcot -e http://idp.tas3.pt:8081/zxididp?o=S 'Discovery Svc' \ http://idp.tas3.pt:8081/zxididp?o=B urn:liberty:disco:2006-08 \ | zxcot -bs /var/zxid/idpdimd
If you want zxididp to support delegation, you would usually configure following registrations as well:
zxcot -e http://idp.tas3.pt:8081/zxididp?o=S 'IDMap Svc' \ http://idp.tas3.pt:8081/zxididp?o=B urn:liberty:ims:2006-08 \ | zxcot -b /var/zxid/idpdimd zxcot -e http://idp.tas3.pt:8081/zxididp?o=S 'People Svc' \ http://idp.tas3.pt:8081/zxididp?o=B urn:liberty:ps:2006-08 \ | zxcot -b /var/zxid/idpdimd
A problem that arises with the bootstraps is the support for recursive or multilevel web service calls. The called service, which receives the bootstrap credential, may legitimately call other services in turn. If it receives the discovery bootstrap as part of its own credential (which may have been a bootstrap in its own right), then it can go and discover further services. The discovery bootstrap does not have to contain any further bootstraps because the discovery service itself knowns about the other services even without receiving a bootstrap for each of them individually.
Now consider attempting to operate without discovery: service bootstraps for first level services need to contain the bootstraps for second level services, which needs to contain bootstraps for third level services. There is no easy way to know how many levels deep the call structure may be. This effectively leads to infinite recursion in bootstrap generation.
Clearly we need some criteria to cap it. In lack of good and logical criteria, we adopt arbitrary one: all bootstraps, except discovery, will only propagate one level, i.e. they will be supplied in SSO assertion, but only the discovery bootstrap will be supplied in the further token assertions.
Fig-2: Recursive bootstrap generation call graph
As described in the metadata registration section, zxidcot.pl web GUI can be used to create discovery registrations. Just point your browser to https://idp.example.com/zxidcot.pl?op=direg and fill in the form fields:
URL for sending SOAP messages
Human readable description of the service
a URN to identify the kind of service offered by the WSP.
The underlying implementation uses the zxcot -e | zxcot -b command line technique.
zxidatsel.pl CGI script provides a web GUI for attribute release consent and selection by the users. This interface can be enabled by setting
You can control the look of Attribute Selection by editing an.css stylesheet and atsel-main.html template.
Usually Attribute Selection is accessed as part of the SSO flow right after authentication step, but it is also possible to access it directly to manage User's attribute sharing preferences. In this respect it serves as a Privacy Manager.
In addition to the official audit trail logging, zxididp logs certain basic events on per user basis. This serves to allow easily displaying to the user recents events about him, raising awareness and allowing user to spot anomalous transactions. The log is kept in file /var/zxid/idpuid/UID/.log
To adapt non- IdPs to environment, the strategy of using SAML2 Proxy IdP profile is recommended. The SP redirects the use to a enabled proxy IdP (aka "middle IdP"), which then offers the user a choice of actual (non-) IdP to use and plays the SAML SP role towards that IdP. When the user has been authenticated, the assertion is returned to the middle IdP, which will use information in it to mint an assertion that is returned to the enabled SP. The assertion SHOULD contain the attributes of the original assertion. It MAY contain the original assertion as well, if audience restriction permits this.
The Proxy IdP Profile can also used for facilitation of interoperation across trust networks. SPs in one trust network use the IdP in their home trust network, which then contacts the foreign IdP. This way only the home trust network's IdP needs to have trust relationship with the foreign IdP. This is much more scalable than each SP having to trust directly the foreing IdPs. See [LibertyInterFed] for further discussion.
The Proxy IdP Profile is described in [SAML2core] section 188.8.131.52 "Proxying" (pp.54-55) and also in [SAML2iop] section 3.3.1 IdP Proxy Feature (pp.11-12), as well as in [SAML2iopDGI] Step D (p.17-19) associated with "IdP Extended" and "SP Extended" conformance modes.
As of version 0.47 (20100114) zxididp supports the following specifications
TAS3 Architecture Core
TAS3 Trust and Privacy Negotiation
Master and TAS3 audit busses
SAWS over SOAP logging (with OpenXDAS events)
OpenXDAS log format
AMQP logging (with OpenXDAS events)
MASTER project event modelling
AAPML (and CARML) support
XACML Authorization Decision Query for SSO
Old session files will accumulate in the filesystem. You should use a cron job to clean them up
touch /var/zxid/idpses; find /var/zxid/idpses -atime +30 | xargs rm -rf
N.B. the touch is needed in the above command line to ensure that the entire directory is not deleted.
zxididp can be configured to support delegation using Identity Mapping Service and People Service. As of 20100922 the People Service implementation is not complete, but is sufficient to support delegation use cases (we are mainly missing the buddy list manipulation).
Invitations and object IDs have separate namespaces. Invitation can be for known object or it can be open. In the open case it can be accepted by existing object, or a new object can be created.
zxididp uses local idpuid/ directory to implement the IdP logins and to store users federations, bootstraps, and attributes.
/var/zxid/ | +-- idpuid/ Local user ID (local login name) to SHA1 mapping | | | +-- JOE/ Local user has directory whose name is the login uid | | | | | +-- .log Log of operations regarding the user | | +-- .ps/ User's People Service Objects and Inviations (i.e. buddy list) | | | | | | | +-- objid | +-- idpinv/ Invitations | | | `-- INVID Invitation object as LDIF
Invitation (identified by SPtoPSRedirectURL) needs to remember
Corresponding user (or object if invitation is directed to specific object)
Permissions, i.e. which sites and web services the delegatee is allowed to access by virtue of invitation. In effect, invitation becomes a permissions bundle that can be revoked without revoking other invitations.
Usage count (when usage >= max usage, invitation is no longer accepted
Invitation is stored in LDIF file at path
Syntax is similar to LDIF (as is the syntax of .at files)
dn: invid=i341343122 invid: i341343122 uid: joedoe psobj: o12321 ps2spredir: https://sp1.tas3.pt/foto/1234 perm: # See below maxusage: 1 usage: 0 starts: 2010-09-22T23:08:55Z expires: 2010-09-29T23:08:56Z
Object needs to remember
federated id of the invitee (filled in once invitee dereferences invite URL)
Permissions, i.e. which sites and web services the delegatee is allowed to access by virtue of his own identity
Invitation IDs that apply
Syntax is similar to LDIF (as is the syntax of .at files)
dn: psobj=o12321,uid=joedoe psobj: o12321 uid: joedoe nodetype: 0 idpnid: F13432 tag: #golden tag: #friend psobjref: o3232 psobjref: o3233 invid: i341343122 invid: i341343123 perm: # See below
ObjectID (psobj) has particlar privacy threat in that several WSCs may
see them and be able to correlate about the user that the object refers
to (see brief discussion in sec 2.1.4 "
We adopt solution where each psobj id issued towards an entity (SP, WSC) is the psobj encrypted (AES-128-CBC) with key consisting of concatenation of secret (/var/zxid/pem/psobj-enc.key) known to ps server only (i.e. the zxididp) and the Entity ID of the entity. See zxid_psobj_enc() and zxid_psobj_dec() in zxidps.c.
Permissions are already in context of delegator and delegatee and need to capture
Whitelist of site EIDs allowed to delegatee
For each site a list of attribute value pairs (in cgi query string syntax). As of 20100922 all of the name value pairs are for future expansion. For example the illustrated localpath will not be intrpreted or enforced.
Syntax (LDIF attribute with dollar separation)
perm: https://sp1.tas3.pt/sp?o=B$localpath=/protected perm: https://sp2.tas3.eu/sp?o=B perm: *$desc=This+is+wild+card+permission.+Use+with+care.
The permissions lines from both invitation and object are used by the discovery and identity mapping services to determine if delegated operations should be permitted.
When an SP starts delegation sequence, it establishes a delegation
record in delegator's (Alice's) user directory, identifying delegated resource,
and then calls AddEntity() embedding the id of the delegation record
in the PStoSPRedirectURL. It is also possible for the SP to simply
encode Alice and the resource in the PStoSPRedirectURL (in symmetrically
encrypted with key only known to SP). The AddEntity() returns
When delegatee (Bob) clicks the invite URL, he lands on PSa and federates by performing SSO using IdPb. He gets redirected to SP using PStoSPRedirectURL. This allows SP to understand which resource is being accessed in delegated way. Bob will usually perform SSO to SP using IdPb. This allows Bob to be permanetly added to white list of delegation record of Alice's resource, so that in future visits Bob can access Alice's resource directly without having to go through the invitation. A record of accessible resource may also be kept on Bob's account. This facilitates Bob's navigation, but the access control decisions must always be based on delegation record on Alice's side.
/var/zxid/ | +-- ses/ Sessions | | | +-- SESID/ Each session has its own directory | +-- user/ Local user accounts (if enabled) | | | +-- SHA1/ Each local user has a directory whose name is SHA1 | | | of the user's NameID (idpnid) and IdP Entity ID | | +-- .mni Information needed by Name ID management | | +-- PS,SHA1 Long term People Service EPR, kept in its own file | | +-- .deleg/ Delegations (invitations) user has issued | | | | | | | +-- DELEG | | | `-- DELEG | | | | | +-- .access/ Invitations user has received (delegations to access other's resources) | | | | | | | +-- ACCESS | | | `-- ACCESS | | | | | `-- .pw User's local password if any (usually none) | | | `-- SHA1b -> ../uid/Sue SHA1 can also be symlink to local account | +-- uid/ Local user ID (local login name) to SHA1 mapping | | | +-- joe -> ../user/SHA1 Symlink points to the real user directory
Alice's delegation records are held in her .deleg subdirectory:
dn: deleg=d122413,idpnid=FAliceatSP deleg: d122413 idpnid: FAliceatSP rsrc: /protected pseid: https://ps.tas3.pt/ps?o=B # Entity ID of Alice's PS (for future membership queries) psepr: ... # Hold long term EPR for PSa for making membership queries diepr: ... # Hold long term EPR for DIa for making web service calls whitelist: FBobatSP whitelist: FCecilatSP psobj: o123 # PS Object IDs of yet unresolved invitations
Bob's delegation hint record (similar to real delegation record, but interpretted differently) are held in his .access subdirectory:
dn: access=d122413,idpnid=FofBobatSP idpnid: FofAliceatSP rsrc: /protected
Since multiple IdPs can be involved, the user directory structure must account for the possibility that Federated ID assigned by one IdP for one user can be assigned by another IdP for different user. This is arranged by hashing together idpnid and the Entity ID of the IdP (see zxid_user_sha1_name()).
There are three methods for Bob to obtain Alice's Discovery EPR (DIa-EPR)
Using PSb, which will ask from IMa (the Liberty Alliance documented way)
Using PSa, which may ask IMa. PSa-EPR is passed from Alice to Bob by the inviting SP
DIa-EPR is passed from Alice to Bob directly by the inviting SP
[IDWSF2Overview] section 2.4 "Cross-principal Identity Service
Access", pp.19-31, describes how Bob can call Alice's service
(n.b. their description has the roles reversed). In that description,
the crucial step of obtaining Alice's Discovery EPR is facilitated by
Bob's People Service (+NOT Alice's People Service!+) using
Here the assumption was that Bob can discover his own PSb and that PSb
knows where IMa is. This latter knowledge is only possible if Alice
had been federated to PSb, which would presume bidirectional
friendship relationship. Such flow is indeed possible within the
PeopleService specification, using
However, if no bidirectional relationship is desired, then the flow does not work. We need a way for Bob to get Alice's discovery EPR with facilitation only by Alice's resources.
Bob will have federated with Alice's PS during invitation, so it
should be possible to direct
In fact, the delegation object could even remember Alice's DIa-EPR for Bob to use. In this sense the resource at SP that Alice delegates access to is in fact the DIa-EPR. While this is simple, there is certain lack of central control. For example, the SP could decide to pass DIa-EPR to Mallory. Basically the model (2) is preferred as it results in special DIa-EPR that expresses the delegation to Bob. Model (3) uses bare DIa-EPR, in effect impersonating Alice rather than conveying genuine delegation relationship.
Copyright (c) 2009-2010 Sampo Kellomäki (email@example.com), All Rights Reserved. Copyright (c) 2006-2009 Symlabs (firstname.lastname@example.org), All Rights Reserved. Author: Sampo Kellomäki (email@example.com)
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.