Authentication and Authorization Considerations for a Multi-tenant Service

Distributed cyberinfrastructure requires users (and machines) to perform some sort of authentication and authorization (together simply known as \emph{auth}). In the early days of computing, authentication was performed with just a username and password combination, and this is still prevalent today. But during the past several years, we have seen an evolution of approaches and protocols for auth: Kerberos, SSH keys, X.509, OpenID, API keys, OAuth, and more. Not surprisingly, there are trade-offs, both technical and social, for each approach. The NSF Science Gateway communities have had to deal with a variety of auth issues. However, most of the early gateways were rather restrictive in their model of access and development. The practice of using community credentials (certificates), a well-intentioned idea to alleviate restrictive access, still posed a barrier to researchers and challenges for security and auditing. And while the web portal-based gateway clients offered users easy access from a browser, both the interface and the back-end functionality were constrained in the flexibility and extensibility they could provide. Designing a well-defined application programming interface (API) to fine-grained, generic gateway services (on secure, hosted cyberinfrastructure), together with an auth approach that has a lower barrier to entry, will hopefully present a more welcoming environment for both users and developers. This paper provides a review and some thoughts on these topics, with a focus on the role of auth between a Science Gateway and a service provider.