Requirements gathering
Communities contacted
The following communities, either represented in AARC consortium or external, in contact with AARC JRA1.1 or other tasks. Contributing to the definition of the AARC plans.
- DARIAH
- CLARIN
- WLCG
- ELIXIR
- PARTHENOS
- ARIADNE
- BioVel
Other stakeholders/target for the questionnaire
- Education/Cloud providers for education
- Libraries
- IdPs
Summary of the answers collected
BioVel
The trust model, as has been initially described in the section 2.1.1 BioVel users need to access data repositories to search, access and upload data, and to access computing services to elaborate the data and then store back the results of their analysis. The BioVel community is leveraging on both internal service providers, for example for the data repositories, and on the European multi disciplinary e-infrastructures, for example to access computing capacity.
The community is accessing a number of heterogeneous service providers, single sign on (SSO) capabilities are fundamental to enable scalable workflows, together with an uniform authorization infrastructure. The workflows requires also to delegate the authorization of one user to a service to access data or perform actions on the user’s behalf.
BioVel foresee also to interact with citizen scientist, therefore some use cases may require the integration with low level of assurance credentials such as social media credentials.
EISCAT
The main use case for authentication and authorization in EISCAT is to grant access to datasets to the institutions/users who are eligible to download the data. Federated AAI could also make easier the accounting for the data usage.
The control over the access to the datasets have been done, so far, using the IP address of the client. But with the extensions of the use base, including additional affiliates, service providers will need to adopt a more sophisticated authorization mechanisms, with a user-by-user granularity.
Also, the logging of who downloads data is not done, meaning there is no way of communicate to the users any new information of problems with the data they have taken. Also, for the reporting to the owners, there is no information taken for what kind of study the data has been downloaded.
The use case here, would be a good way for authentication of who and possibly why they download data. An 'EISCAT' certificate for users, including who, why, when, how the user will handle the data. One could think of different levels of the certificate for different levels of data.
Currently EISCAT is not using AAI solutions directly integrated with the services.
CLARIN
CLARIN established a task force for the AAI integration in 2013, and the community is deploying AAI integrated solutions. In most countries the IdP systems of the universities and other research institutions offer a high level of traceability of users which is very relevant for allowing access to restricted corpora.
Another obvious benefit is the single sign on, lack of local accounts, increasing password security for the users and making user management easier for service providers.
CLARIN is following actively developments in the AAI field e.g., by establishing and maintaining CLARIN Service Provider Federation.
The CLARIN infrastructure includes about 1000 SPs and IdPs from 220 institutions in 33 countries. Potentially the user base could be 100.000 users (students and researchers of humanities).
For homeless users CLARIN has deployed a community IdP.
In general the experience of CLARIN with FIM is Good within a National Federation, limited in trans national federations (IdP/SPs do not trust each other, insufficient attribute release, no description of the participants).
Most IdPs offer decent login pages, SAML error messages are often very cryptic and give the wrong impression on what side (client or server) the root cause of the problem is, and the services and tools are not implemented to deal in a transparent way with missing attributes.
EGI
EGI is a resource infrastructure. The main goal is to enable users to access the distributed infrastructures.
The process to access EGI services is illustrated in the above picture:
The user, on the left, obtains a personal certificate from a certification authority recognized by EGI, then the user adds the information about their VO to the certificate generating a X509 proxy. To be able to do so the user must be authorized by the VO manager, to simplify the PI of their collaboration, who controls the membership of the VO they are managing.
All the needed information (user identity, Vo memnership, additional roles and capabilities within their VO) are shipped to the EGI services that use these information to authorize the access to the services. Generally access to services is regulated by VO membership: a service provider supports a number of VOs and users can search for the services enabled for their VO. Finer grained user authorization is possible but not often applied since the scale of the infrastructure. The AuthN/AuthZ process is based on the trust between CAs federation and the service providers federation, plus the trust between the service providers and the user communities (VO).
X509 authentication has proven to be scalable and to work for almost any use case, from a technical point of view. New user communities prefer to use other technologies for the authentication, for example username/password based authentication.
EUDAT
Federated access has been one of the goals from the beginning of the EUDAT project. EUDAT infrastructure is designing a new service called B2ACCESS to enable the integration of federated IdP with the EUDAT services. The use cases for this service are:
- Integrating new IdPs and enabling SSO for the users using credentials they already own
- Providing an EUDAT unique identifier for the users
- Providing additional attributes associated to the user identity
- Providing catch-all IdP for homeless users
This solution may support also non federated IdPs, nevertheless it would benefit from federated identity management relying on the LoA and the other best practices implemented in a federation.
As technical solution for B2ACCESS EUDAT is deploying unity-idm.
The level of information available is different in each of the communities and needs to be tackled case by case as more and more communities join the EUDAT infrastructure.
In general the EUDAT experience with AAI solutions is good both in terms of usability and quality delivered. The technologies currently enabled in EUDAT are X509 certificates, SAML2 and OpenID Connect/OAuth2.
B2ACCESS is not yet integrated in eduGAIN, but this is high in the list of priorities, currently it is under testing with an handful of IdPs and users.
FMI
FMI would support many applications cross institutions and countries, but currently the academic AAI have issues in cross-country interoperability, and interoperability – for example – with social media credentials. FMI uses a commercial AAI solution, Auth0, and it works very well for the FMI use case, one example is the integration with a commercial service such as dropbox, which was very easy and fast.
The experience with academic IdPs has been difficult because of the bureaucracy involved, and the limitations for attribute sharing.
D4Science
The main benefits in FIM highlighted by D4Science are:
The easiness in taking part in several research communities by using a single set of credentials.
Less barriers for users in accessing services exposed by different research communities.
Adoption in conjunction with Access Delegation, Federated access enables capability to build composite services built on top of services spanning across different identity domains.
In general using using the SAML Web Browser SSO profile tools are user friendly and they fulfill the user expectations, where such expectations are SSO capabilities with a adequate level of security.
Adopted Authentication & Authorisation Technologies
BioVel
The community is still at the beginning of adopting federated identity/authorization solutions. Working closely with EGI and other service providers using X509 certificates as an authentication mean, the community is relying on the IGTF certification authorities federation. The overhead of obtaining and maintaining a personal certificate could though be seen as an excessive overhead by many new users. This solution is also not feasible for homeless users.
DARIAH
The DARIAH infrastructure blocks are built within national initiatives. AAI is based on SAML authentication combined with attribute aggregation. A DARIAH homeless account is available.
Personal data of users are stored in a central clustered LDAP server. Group memberships that provide access to services and Wiki spaces, as well as the user data are managed via a web-based administration portal. Attribute queries, as defined in SAML and implemented in Shibboleth, are used to aggregate information from the campus IdP and the DARIAH Attribute Authority implemented in the DARIAH IdP. A registration mechanism based on a central DARIAH SP ensures that all personal data that are are needed, but not provided by the Campus IdPs, are collected as self-asserted data from the user. The DARIAH IdP thus acts as an IdP-AA, but not as an SP, i.e. it is not a proxy.
ELIXIR
The overall ELIXIR architecture is split into three main layers. The bottom layer comprises external authentication providers and serves as a source of identities to the middle layer, i.e. the ELIXIR AAI layer. More specifically, the ELIXIR AAI layer consists of the ELIXIR Directory as a main storage of user data, related services to manage the user data, and services to authenticate the user. The upper layer is composed of relying services for the ELIXIR users. Those relying services can utilise data from the ELIXIR directory. A brief overview of each component follows.
External authentication providers authenticate an end-user and deliver their authenticated (external) identity to the ELIXIR AAI. ELIXIR supports both Home IdPs (e.g. via eduGAIN inter-federation service) and Common IdPs (Google, LinkedIn and ORCID).
The ELIXIR Proxy IdP is an authentication proxy that authenticates an end-user using an external IdP or their internal ELIXIR password (if they don’t have an external identity), and relays the authenticated identity and associated attributes to the Relying services. The ELIXIR Proxy only passes authenticated users’ attributes to the Relying services for authorisation. Associated with the ELIXIR Proxy IdP there is also a service where an end-user can consolidate his/her external identities and the ELIXIR identity, by logging in using the ELIXIR password and/or two or more external IdPs consecutively. The ELIXIR Proxy IdP can also support step-up authentication or credential translation.
The ELIXIR Directory is a central storage which stores identity and attribute information on each ELIXIR user. ELIXIR users can use the attribute self-management service where they can self-manage some of their own attributes (such as preferred language and ELIXIR password), register other identifiers (ORCID, social ID), change their ELIXIR nickname associated to the attributes in the ELIXIR Directory.
ELIXIR must ensure to provide some of its services only to the users with an affiliation with academic institution. The ELIXIR AAI will provide a bona fide based status application for bioinformatics researchers, so that the community can vouch for them.
The ELIXIR group management puts strong emphasis on delegation of rights. Group managers can add users to groups via invitation, directly and, in the future, by importing group information from external systems. All group information is exposed to the ELIXIR Directory.
The Dataset authorisation management component is used by ELIXIR users where they can apply for, and dataset owners approve, access rights to datasets. The access rights are exported to the ELIXIR Directory and/or directly to the Relying service that enforces the access rights.
EGI
EGI is a highly distributed infrastructure, multi-disciplinary, integrating more than 300 resource centres (service providers) and almost 20,000 users grouped in 200 user communities called Virtual Organizations (VO). Currently, authentication and authorisation within EGI is enabled through a X509-based Public Key Infrastructure (PKIX), based on the Interoperable Global Trust Federation (IGTF) and EUGridPMA Certification Authorities federation.
X509 technology, and in paritcular X509 proxies, enables some of the most important capabilities for EGI: scalability, command line access and delegation. The user actions are usually initiated by submitting a request with attached an X509 proxy, this means that in case of bulk submissions of - as an example - thousands of computing tasks, the services do not need to contact the IdP or the attribute authority of the users thousands times, since all the needed information are in the proxy. The X509 credentials work with no issues with command line commands, and the proxy implement a form of delegation (impersonation) that allows a service to perform a defined set of actions on behalf of the user.
From a technical point of view, X509 authentication has proven to be scalable and to work for almost any use case. However, new user communities prefer to use other technologies for the authentication, for example username/password based authentication.
There is a capillar network of certification authorities and registration authorities, distributed among the EGI partners, which can be contacted by users to obtain a certificate. EGI runs a catch-all CA to support users who - for any reason - cannot access an existing CA.
To bridge different authentication technologies with X509, the EGI partners and the user communities are deploying science gateways and portals where users can authenticate with username/password, and access the resources through web-based tools and interfaces. The portals are then generating short-lived X509 credentials that are used to access resources.
The most common mechanism used to bridge between IdPs and X509 are the robot certificates, which can generate programmatically short lived X509 proxies that are used to access EGI services. One of the drawbacks of this solution is that the real user identity is hidden behind the robot certificate. To partially address this issue, EGI is implementing an extension of the X509 proxy certificate that contains an ID that can identify the user if needed. This is particularly useful for accounting purposes and, at the same time, improves the overall security of the implementation.
EUDAT
The EUDAT infrastructure is implementing an AAI solution called B2ACCESS which bridges various AAI technologies used by the communities with the various technologies used by the service providers within the infrastructure. Users of the infrastructure must be able to use their home organisations’ IdPs to log in to the EUDAT infrastructure, as well as IdPs run by the user communities (e.g. CLARIN) and, when applicable, social media (e.g. Google). The user’s EUDAT identity contains attributes provided by the “home” (external) IdP, supplemented with attributes maintained by B2ACESS itself (the proxy scenario) - so every internal identity can present, for example, an email attribute. Finally, EUDAT itself can also provide a “homeless” IdP by having users authenticate using username/password.
As EUDAT runs a lot of different services internally, SPs are offered different technologies for integration with B2ACCESS. Based on work originally done in the Contrail project and carried forward by the first EUDAT project, services can choose to use SAML (using the WebSSO profile), OAuth2 (RFC 6749), or X.509 certificates. In a simple scenario, a service will use OAuth for simple access and authorisation; a more complex scenario would use SAML to authenticate and pass authorisation decisions to an XACML infrastructure. In the most elaborate scenario, users would authenticate to a (community or EUDAT) portal using SAML WebSSO (which in turn redirects to the “home” IdP); then the portal obtains an OAuth access token to grant further rights, including the right to obtain a certificate on behalf of the user from a web services certification authority (internal to B2ACCESS) - it then generates the keys and obtains the certificate which includes authorisation attributes. This scenario provides extra security compared to many of the traditional portals as the issuance of the certificate requires additional authorisation (the portal has to be registered as an OAuth client and needs to be authorised to obtain a particular user’s certificate); the management of such authorisations can be predefined administratively or by the user and can be audited subsequently. Power users can download the certificate and use it to drive non-web services from the command line, e.g. B2SAFE and data transfers. Despite its complexity, this approach offers a high degree of flexibility of integration of the many different components that comprise an EUDAT infrastructure, and many services will choose to use only the least complicated AAI mechanisms depending on their security requirements.
D4Science
The Authentication Service supports username/password and X509 Certificates factors: the former is used mainly for human-machines interactions (Portal or SOAP Services) and the latter is mainly used for machine-machine interactions (SOAP Services called by other services).
The user interface portal support SAML federation with the classic Web Browser SSO profile. An hybrid profile based on Web Browser SSO and SAML Token profile enables SAML federation for SOAP Web Services.
FMI
FMI is using Auth0 (https://auth0.com/), which is a commercial solution for AAI integration.
Education
The institutional IdPs are using SAML2 technology, and solutions like Grouper for the attribute management.
Penetration of federated identity management
BioVel
Although BioVel management understands the need for federated identity management, the community has limited experience with federated AAI solutions. The research community has not already an AAI solution for the community in place, and therefore there is still need to acquire the needed knowledge.
Most of the users have credentials from their institutional IdPs, but the percentage of these federated in eduGAIN or other federations has not been assessed.
DARIAH
DFN-AAI/eduGAIN is feasible and being used by a number of users. However, there are lots of user accounts in the homeless IdP LDAP server for users that either have no federated IdP or with an IdP that does not release ePPN
And there is some number of users that simply are aware that "a DARIAH account" can be their institutional account, who even do not try to log in via AAI, going for the homeless user option.
PSNC
In Poland there are not many institutions federated with eduGAIN or PIONIER.Id, therefore an important fraction of the users have no access to federated ids.
EGI
EGI users are mostly using credentials released by the IGTF federation. the X509 technology is fulfilling the diverse use cases of the EGI community. Never the less EGI also has many use cases of users using federated identities, there is not direct integration with the EGI resources, but science gateways and other web user interfaces have been successfully integrated with eduGAIN federated IdPs or in general with other non-IGTF IdPs.
In summary at the moment of writing EGI users can use - if needed - credentials from federation different from IGTF, but not being directly integrated in the resources, this feature must be implemented differently for every use case.
EGI strategy is to push forward the integration with federated identity providers directly in the services. The expectations are to have direct SSO support for the operational tools that enable federation and cloud services. HTC services will likely leverage on translation mechanisms to X509.
To implement this integration, a uniform policy of attribute release and automatic management of the Level of Assurance (LoA) would be beneficial.
EUDAT
EUDAT expects eduGAIN to cover most of the user base of the infrastructure. The B2ACCESS service will enable connection with other IdPs or federations upon request of the users. EUDAT is developing training material for the infrastructure services, including B2ACCESS.
FMI
The FMI users are often not federated in national federations, for example hospitals or commercial companies, as a consequence the service provider needs to federate with the individual IdPs.
Education
At the moment of the assessment 42 universities out of 80 (Italy) had an institutional IdP in place and running. Currently there are approximately 4 millions of users, out of the 8-10 millions that can potentially join with a 100% coverage.
Perceived barriers
BioVel
Currently the AAI federations, and AAI coordination activities, have been focusing on the end-user to Service provider direct interaction, in other words enabling simple SSO capabilities on the services. This approach is necessary but not sufficient to fulfil (probably) the BioVel requirements, which envisage a more complex relation between service providers, which need to interact to enable the workflows of the community. Delegation and uniform authorization across service providers will be fundamental bricks of the BioVel infrastructure.
The main barrier for BioVel is the lack of information and knowledge, and the community would benefit from a reliable and organized source of information, in form of online documentation, which can be consulted to take informed decision. Possibly integrate with trainings. Currently the information is very scattered and it is challenging to get the full picture that includes the IdPs documentation, the IdPs federations and the SP federations requirements and best practices.
The training and the documentation should be integrated with a support service and troubleshooting tools, to maximise the efficiency of the federations.
EISCAT
In general the EISCAT community lacks information about federated identity management.
EGI
Being EGI an highly distributed infrastructure, with hundreds of service providers, the lack of scalable policies for the release of the attributes from the IdPs is definitely a show-stopper. With hundreds of user communities (thousands of users) and service providers, integrating potentially hundreds of IdPs with all the service providers would require an excessive amount of effort, if possible at all.
The use cases of EGI have technical requirements that require command line access and delegation capabilities, which are not extensively supported by all the AAI federations.
From an internal point of view, the main barrier is that a good fraction of the legacy software used for resource provisioning will never be ported to technologies dfifferent from X509, therefore credential translation is the only way to access those services.
EUDAT
The main barriers for a wider coverage of IdP federations among communities and service providers from the EUDAT point of view are the lack of technical knowledge and sometimes of resources. Also the bureaucracy involved in joining a federation as an IdP or an SP is considered a barrier based on the EUDAT experience.
D4Science
The main barriers to expand the integration with federated identity management are in some cases the lack of resources/funds and the bureaucracy overhead involved in joining a federation. It is not always clear, for all the use cases, if the benefits of joining a federation compensate the overhead.
FMI
The main barrier perceived is the excess of bureaucracy both to use an IdPs as a service provider and to join and IdP federation. Another problem is the lack of flexibility in the academic IdPs, in terms of technologies supported or willingness to support new use cases.
Based on the FMI experience, it is very difficult to join national AAI federations in some countries, where for example commercial providers cannot join due to NREN policies. On the other hand a cross-national common AAI provider would solve most of the integration issues currently experienced, similar to the providers of X509 host certificates.
Education
Based on the summary survey provided by GARR the main barrier in adopting federated identity management for the universities and education institutions are the following:
First barrier is the low priority that FIM has in the management agenda of many campuses and educational institutions. All the other barriers are a consequence of the fact that management does not consider AAI important: lack of funds and expertise, for example. Lack of strategy, for example should IdP be a campus service or an externally provided service?
Requirements: Authentication and authorization technologies
The stakeholders invovled in the survey in some cases have been already using AAI technologies, and may have requirements associated to this topic: type of IdPs, standards to be considered in the AARC architecture, need for delegation technologies or non-web access to services.
BioVel
BioVel community will move in the direction of using the institutional credentials, possibly federated in eduGAIN, to access the services specific for the community but also possibly to access the multi-disciplinary e-infrastructures. In other words, institutionally affiliated login Shibboleth/SAML federated plus a solution for homeless users.
Support for delegation is a complex matter under investigation in BioVeL. Chains of delegation are needed, both for authentication information and authorization information, and ultimately for accounting purposes.
DARIAH
DARIAH user authentication is leveraging on the institutional IdP of their users, part of national federations such as eduGAIN federations, and the catch-all community IdP to host homeless users, who are a consistent fraction of the community user-base.
DARIAH is interested in SAML2, and OpenID/OAuth2 technologies, plus X509 credentials for legacy reasons.
It is important for the community that the authentication technologies are as much user friendly as possible. For the community is also important the support for delegation and non-web access, on top of the normal web accessible services.
PSNC
Users of the PSNC services need to use their institutional IdPs to access the services, depending on the service offered, also catch-all community IdP or social media can be relevant. User friendly technologies to access web based an non browser based services are important.
Photon and Neutron
The photon and neutron community is using the Umbrella infrastructure for authentication and authorization. The technologies relevant for the community are SAML2 and X509 certificates.
Umbrella users are using credentials from their institutional IdPs and the federation in eduGAIN is an added value, alternatively users who cannot access to a federated IdP can use the catch-all IdP provided by Umbrella itself.
For the photon and neutron use cases AAI must support: easy single sign on solution, web based and non-web applications, delegation.
EGI
EGI authentication is based on X509 certificates, released by the IGTF federation. From a technical point of view X509 credentials satisfy most of the requirements of the EGI infrastructure.
Authorization on EGI services is mostly regulated by the membership to a virtual organization (VO). Services support VOs, and give access to the resources to the members of the VO. Finer levels of granulaity are possible with user groupings or roles within the VO. All these information - which are in fact community attributes - are added as extensions to the proxy certificate by the VOMS service (Virtual Organization Management Service). VOMS allows the VO Managers, the community responsible, to manage autonomously user membership and the other attributes. In this way EGI service providers delegate authorization to the communities.
EUDAT
For a multidisciplinary infrastructure as EUDAT it is important to keep open the possibilities for the users to choose the IdP of their preference, this includes institutional IdPs, both federated in national federations and non federated, catch-all IdP provided bu the research communities, and social IdP. The LoA associated to them is also important.
Within the EUDAT infrastructure many services are offered and each service has it’s own needs with respect to authentication technologies. Currently EUDAT is supporting X.509, SAML and OAuth2. Some services are web based but others are not. In this case the infrastructure is currently using X.509 based solutions. Moreover, EUDAT aims for a single sign on experience but given the complexity of the infrastructure this is not always possible.
FMI
The best solution for FMI would be to get AAI as a service, as provided by the commercial provider currently used. FMI needs to integrate IdPs both federated and non federated in the national federations, as well as social IdPs for the homeless users.
Easy and user friendly SSO, friendly for the end users but also for the service providers, support for both web based applications, and command line applications.
D4Science
D4Science services are enabled for SAML2, X509 and (under evaluation) Open ID Connect, OAuth2, in terms of sources of identity information, the infrastructure will need to access institutional IdPs, catch-all IdPs managed by the communities and in the future social credentials.
The federation with the national federation is under evaluation, and in general D4Science would benefit of simpler procedures to join an IdP federation as service provider.
D4Science services would benefit from delegation and non-web access capabilities.
CLARIN
CLARIN users are using, or will use, both their institutional credentials federated and not federated in national federations. Plus the CLARIN catch-all IdP, in order – if possible – to avoid the use of social credentials.
CLARIN use cases involves both web-based and non-web based access to services, and also delegation capabilities, which are being currently tested.
Training on how to deal with the shortcomings of FIM would be helpful: SAML error messages, testing connections, how to request attributes in foreign federations. CLARIN addresses some of the issues via the Service Provider Federation. We also attempt to address these issues ourselves, but most of them are generic FIM issues.
There should be more information for IdP operators on how SPs use them and what problems they have to solve.
Education
For the education community the institutional IdPs are the most likely credentials that the users will use. Governmental IdPs, e-ID, could be also potentially interesting once they will reach full maturity.
Either solution will be predominant, user friendliness is the higher requirements for this community.
Currently the coverage and the features of the national and European IdP federation are not entirely fulfilling the needs of the education sector. Besides the campuses not yet federated there are other categories of users who would be left out anyway: for example the high school students not yet registered in a university but who need to access some tools, and the citizen users in the university library.
Requirements: Attribute release policies