170 likes | 188 Views
Explore the scope of work, architectures, policies, and technologies of Shibboleth and Web-ISO for web resource sharing and user authentication. Learn when to use each system and the potential for long-term convergence.
E N D
Shibbolethand Local Web-ISO Scott Cantor cantor.2@osu.edu The Ohio State University Internet2/MACE
Shibboleth “Scope of Work” Architectures, policy structures, practical technologies, and an open source implementation to support inter-institutional sharing of web resources subject to access controls. • Federated Administration • Access Control Based On Attributes • Active Management of Privacy • Standards Based • Extensible Policy and Trust Framework (“Clubs”)
Web-ISO “Scope of Work” Share expertise and develop common solutions to standardize user authentication to intra-enterprise web-based services with single sign-on and other useful functionality. • Multiple domains, but not multiple organizations • Strictly deals with authentication and identity • Protects privacy only externally to system • Collects requirements from many proprietary SSO systems
Shibboleth “fits” with Web-ISO • Implements an industry-standard SSO protocol designed for both inter- and intra-organizational use • Capable of communicating a superset of the information Web-ISO provides to applications • Applications used by extra-organizational users are also used by intra-organizational users • Other application and security requirements are equally applicable to both systems
Shib “misfits” w/ Web-ISO • Pubcookie (and other code bases) are more mature in performance, stability, and features • Focus of Shibboleth architectural work is on the new features beyond Web-ISO • Privacy features imply that obtaining username requires an extra lookup • Security properties of the SAML SSO profiles may not match that of some Web-ISO implementations
When Shibboleth Makes Sense • Deployed Web-ISO is primitive, non-existent, or aligns well with Shibboleth design • Enterprise LDAP and/or authorization services are immature or under-supported relative to web security projects • Internal politics create information boundaries that privacy/trust controls can help to cross • Simplicity of application access to attributes is a virtue
When Web-ISO Makes Sense • Existing approach aligns better with Web-ISO design and assumptions • Advanced features of system are deemed important in the short term • Mature directory and/or authorization services are available for use by applications • Libraries readily available to simplify efficient access to attributes (caching, redundancy, etc.) • Majority of services want username and little else
Long Term Convergence ofShibboleth and Web-ISO Assuming SAML maintains momentum, the Web-ISO project can meet requirements for better standardization by moving toward it. A SAML-compliant Web-ISO fundamentally differs from Shibboleth only with respect to trust/policy requirements and privacy controls. Timetables and plans for such convergence are not currently in place.
Short Term: Using Shibboleth with Web-ISO • Shibboleth Handle Service can be deployed as a locally-secured application • Attribute Authority might utilize similar infrastructure to that used by applications • Applications straddling internal/external boundary must accommodate both modes of access • Some overlap in management of the two systems
Short Term: Using Shibboleth as Web-ISO • Somewhat ahead of the project, so some gap-filling required • Handle Service expected for Beta-1 should provide cookie-based SSO across a replicated array of login sites, but still assumes authentication is already done • Small amount of work to merge in an existing form-based login service • PKI-based server interactions may be new
Short Term: Using Shibboleth as Web-ISO • Analysis of required target-side functionality • Platform support (Linux, Solaris, Windows), but basic portability should be very high • Web server support (Apache 1.3), IIS likely to be next priority, but OSU (at least) will make iPlanet and Apache 2 versions available quickly • Performance, performance, performance
Shibboleth at OSU:A Case Study • Homegrown Web-ISO assigns each session a UUID, then targets query back for user attributes (username, SSN, employee ID) via secure RPC and export attributes to HTTP headers for use by applications or access control modules • Shibboleth assigns each session a handle, then targets query back for user attributes via SSL, and export attributes to HTTP headers for use by applications or access control modules Hmm…
Shibboleth at OSU:A Case Study What’s Different: • DCE-RPC is a lot faster than XML over SSL • Management of server identity and attribute access currently easy with DCE and groups; Shibboleth introduces PKI and ARPs • Some targets currently don’t have an SSL key-pair, where do they get one for free? • Should the entire local deployment be “club-compliant” or should there be a wall between external and internal Shibboleth?
Shibboleth at OSU:A Case Study What’s Different: • Most important attributes used by OSU other than EPPN not yet defined by club • Existing applications may expect usernames of a certain format or length (“legacy mode”), or may conflate authentication with authorization (gasp!) • Existing Web-ISO does have some features not currently in Shibboleth code base (but not many) • Platform support (IIS, iPlanet, Apache)
Shibboleth at OSU:Tentative Plan • Finalize SOW for Shibboleth 1.0 (what’s in/what’s out), to identify expected feature gaps • Fork Shibboleth code at OSU to add missing features, documenting the extensions for possible broader adoption (Liberty?) • First decision point: is LDAP service “cooked” enough or will relational data store be necessary?
Shibboleth at OSU:Tentative Plan Phase One: • Deploy Handle Service behind existing Web-ISO • Migrate selected target servers (starting with mine) to Shibboleth for evaluation, in “legacy mode” • Should allow simultaneous SSO across both systems • Final acceptance testing by cutting over high volume applications at a busy time (grades, scheduling)
Shibboleth at OSU:Tentative Plan Phase Two: • Deploy redundant Handle Services as stand-alone authenticators • Migrate remaining servers • Shut down legacy system! • Applications can be upgraded in parallel to decrease use of “legacy mode”