Difference between revisions of "Projects/VerifyAuthData"
(→Architecture) |
(→Architecture) |
||
| Line 16: | Line 16: | ||
A new opaque type, an authorization data context (krb5_authdata_context) provides an attribute-based accessor onto a set of authorization data elements. The operations which one can perform on this context roughly mirror those specified in draft-ietf-kitten-gssapi-naming-exts. |
A new opaque type, an authorization data context (krb5_authdata_context) provides an attribute-based accessor onto a set of authorization data elements. The operations which one can perform on this context roughly mirror those specified in draft-ietf-kitten-gssapi-naming-exts. |
||
| − | An authorization data context is always created and authenticated by krb5_rd_req(). Access to it is provided via the krb5_auth_con_get_authdata_context() API. (Note this function returns a weak reference; it is freed along with the auth context.) |
+ | An authorization data context is always created and authenticated by krb5_rd_req(). Access to it is provided via the krb5_auth_con_get_authdata_context() API. (Note this function returns a weak reference; it is freed along with the auth context. Keep your reference along side the auth context, and copy it if you need another one.) |
The authorization data context encapsulates a set of authorization data <i>systems</i>, each which provide support for one or more authorization data types. Each system is instantiated as a <i>module</i>, one for each authorization data type. This is modelled closely on the client-side preauthentication SPI. The set of systems is extensible at runtime through a plug-in interface. |
The authorization data context encapsulates a set of authorization data <i>systems</i>, each which provide support for one or more authorization data types. Each system is instantiated as a <i>module</i>, one for each authorization data type. This is modelled closely on the client-side preauthentication SPI. The set of systems is extensible at runtime through a plug-in interface. |
||
Revision as of 13:24, 25 August 2009
Background
The goals of this project are to:
- change the behaviour of krb5_rd_req() to always verify known authorization data elements
- provide an attribute-based GSS interface for extracting authorization information (draft-ietf-kitten-gssapi-naming-exts)
- provide a vendor-extensible plug-in interface for verifying and extracting authorization data elements
- eventually, use this interface to allow authorization data elements to be submitted in KDC and AP requests
Architecture
A new opaque type, an authorization data context (krb5_authdata_context) provides an attribute-based accessor onto a set of authorization data elements. The operations which one can perform on this context roughly mirror those specified in draft-ietf-kitten-gssapi-naming-exts.
An authorization data context is always created and authenticated by krb5_rd_req(). Access to it is provided via the krb5_auth_con_get_authdata_context() API. (Note this function returns a weak reference; it is freed along with the auth context. Keep your reference along side the auth context, and copy it if you need another one.)
The authorization data context encapsulates a set of authorization data systems, each which provide support for one or more authorization data types. Each system is instantiated as a module, one for each authorization data type. This is modelled closely on the client-side preauthentication SPI. The set of systems is extensible at runtime through a plug-in interface.
An authorization data system for the Windows 2000 PAC is built into libkrb5, as a wrapper around the krb5_pac APIs introduced in MIT Kerberos 1.7. Plug-ins have preference over internal systems, so a vendor that wishes to expose inner PAC fields (such as SIDs) as attributes could do so by registering the appropriate plug-in. (In the PAC case, this would avoid the overhead of building a NDR parser into libkrb5. Moreover, it is a generalised interface that should encourage the use of authorization data for other purposes.)
This interface is not designed to be used for authorization data construction or verification within the KDC, at least at this stage. The existing authorization data SPI contributed by Apple and Novell should be used for this.
Implementation
libkrb5
GSS
Example
% ./t_namingexts test.keytab Target name: host/test.de.padl.com@DE.PADL.COM init module "mspac", ad_type 128, flags 00000004 init module "mspac", ad_type 128, flags 00000004 Source name: host/test.de.padl.com@DE.PADL.COM Attribute mspac: Authenticated Complete 050000000000000001000000b001000058000000000000000a00000034000000 ... Attribute mspac:logon-info Authenticated Complete 01100800cccccccca001000000000000000002002caaea1ebc24ca01ffffffff ... Attribute mspac:client-info Authenticated Complete 8089662f7925ca012a0068006f00730074002f00720061006e0064002e006400 ... Attribute mspac:upn-dns-info Authenticated Complete 2a00100016004000000000000000000068006f00730074002f00720061006e00 ... Attribute mspac:server-checksum Authenticated Complete 10000000ff41c35408ef41128530200a Attribute mspac:privsvr-checksum Authenticated Complete 76ffffffdfb93706e991880236aa70130e1e87a9
Status
Code is in the users/lhoward/authdata branch.
