Internet2 PKI Labs

Draft in progress, Oct 2000.

General Issues

• Secure-computation for certificate path discovery when directory information is not public, including delegation of specific authorizations from within an organization
• Securing trusted computing bases and signing environments to provide a meaningful level of non-repudiation without significant legal risks to the keyholder.
• PKI-enabled Workflow system with Authorization certificates and delegation
• PKI Tools for Pervasive Computing
• Namespace conflicts between CAs which are cross-certified or bridged in the PKIX model.
• Integration of PKI with the Grid: Condor and Globus Security
• A Language for Authorization and Policy Management
• Hardening PKI with Secure Coprocessors
• Minimizing the privacy risks that can accompany wider use of certificates

Human Interface Issues

• Building a Secure, Usable Browser. Securely showing the user whether or not the keyholder serving up the page being viewed is allowed to use a particular logo or trademarked word found on that page when a web site outsources the e-commerce function. Finding a secure subset of javascript.
• email user interfaces that minimize the "John Wilson Problem". As directories get more comprehensive users are lured into sending sensitive email to their best guess of the right person, resulting in too many dangerous mistakes. Names work for humans only when the set of names is small. The Internet is not a small community.
• Overcoming the Inconvenience Barrier to PKI

Does support for pseudonymity increase or decrease software or system complexity, and if so to what degree....

Draft Research Agenda

1. General issues
   1. Addressing problems of scale for PKIs in large communities of interest composed of relatively autonomous entities, like Internet2 itself.
   2. Alternatives to CRLs. A number of alternatives to CRLs are being developed to guarantee the currency of certificate data. The relative advantages of these approaches are not well understood, nor are testbeds available to evaluate these options.
   3. Interoperability between systems and infrastructures based on various approaches (PKIX, Kerberos, PGP, SPKI, GSS-API; different public key algorithms; platforms; etc.)
   4. Optimizing architectures and infrastructures for different constraints: simplicity, security, privacy, cost, performance, connectivity, availability, legal environment, various application scenarios. E.g. tracking down referral chains between certificates can be slow.
   5. Defining and measuring metrics for PKI: performance, availability, etc.
   6. Usability and human interface analysis. Human interface techniques for dealing with issues like multiple valid certificate chains which have differing policy implications or confidence levels.
   7. Methods to support the simple and secure installation, maintenance and update of trusted CA root certificates and security policies to clients, and facilitating (when appropriate) the sharing of security-related settings across various end-user devices.
   8. Securely moving expensive or complex client-side processing or connectivity to nearby computing and networking resources (certificate chain resolvers, ldap referral chaining, etc.)
   9. Identification and/or mitigation of security threats specific to PKI-related protocols and services: flaws in relevant protocols, Denial of Service attacks, compromise of CA private keys, audits of code, certificates and policies, etc.
   10. Harmonizing approaches to Single Sign On with PKI systems.
   11. Minimizing the privacy risks that can accompany wider use of authentication.
   12. Maintaining access to secret keys and the data they protect despite forgetful or unavailable key owners.
   13. Gradated approaches to authorization and authentication in which the level of certainty is communicated to the application, so it can be balanced against the cost of an error, and additional evidence can be requested.
   14. Name-space management issues.
2. Authorization Infrastructures
   1. Public Key Infrastructures based on attribute certificates vs. identity certificates.
   2. Approaches to delegation of authority.
3. Policy Issues
   1. Use of policy objects with certificates, and the management (construction and validation) of complex trust chains (varying over context and time). Also, the ability to do effective policy mapping between disparate PKI domains issuing certificates under different certificate policies.
   2. Policy definition languages
   3. Automated generation of certificates, incorporating information from existing sources and user input according to a policy
4. Related Directory Services
   1. Development of uniform access control model for access to directories in general, and to specific information within those directories (granular control).
   2. Analysis of the performance of chaining versus referrals for the finding and processing of trust paths.
5. Open Source Solutions
   1. New libraries, modules, plug-ins, applications, user interfaces, etc., in open source code suited to a variety of platforms and languages.
   2. APIs and configuration systems to hide the complexity and diversity of security infrastructures from the applications programmer.
   3. Prototype implementations of common applications (inter-library loan, research collaboration, shared access to preprint servers, IPsec, etc.)