Release Meeting Minutes/2013-03-26
From K5Wiki
Shawn Emery, Will Fiveash, Thomas Hardjono, Greg Hudson, Ben Kaduk, Eric Kozlowski, Simo Sorce, Zhanna Tsitkov, Tom Yu
- Shawn
- Tom, extra-round-trip draft?
- Tom
- Haven't looked in detail.
- Tom
- Solaris -- one function per auditable event type.
- Shawn
- Event elements not changing that much... last change was maybe 5 years ago. JSON allows flexibility but events won'd change that much. Authorization data might be more volatile. Not sure what other vendors want. Maybe audit authorization data.
- Tom
- That depends on what customer requirements are.
- Tom
- Authorization data can't be generically decoded. Not necessarily ASN.1 encoding.
- Simo
- MS-PAC?
- Shawn
- Ticket flags, etc.
- Tom
- Authorization data... nesting of containers
- Simo
- Maybe authorization data auditing in the authorization data plugins. Defining events on the fly.... useful for plugin to add its own auditing.
- Greg
- How?
- Simo
- Key-value interface. Plugins generate events. e.g. how plugin constructs authorization data.
- Will
- Separate library that plugins can link to? Other plugins e.g. authorization data to audit.
- Simo
- Some people would like to see authorization data being generated.
- Greg
- Per-event methods would have access to entire req/rep. Dmitri wanted key/value pairs to have flat numeric or string values. Hard to do that for authorization data.
- Shawn
- Simo, audit on application server side?
- Simo
- We might to some degree.
- Zhanna
- You said events not changed for last 5 years. API stability? not needed? etc?
- Shawn
- Looking at design 1. Components have been around for ~5 years, e.g. S4U has been here ~5 years (since 1.7).
- Tom
- No harm in logging flags numerically, because backward compat etc.
- Shawn
- Common Criteria requirements for authorization data?
- Zhanna
- Not necessarily. They have preauthentication
- Thomas
- Verify correctness of policies?
- Shawn
- Uninterpreted authorization data... would log as blob.
- Tom
- Simo's idea re plugins making own audit events.
Shawn talks about postprocessing authorization data binary blobs.
- Simo
- What if something in blob shouldn't be logged? Authorization data plugin would have better idea of what's sensitive.
- Zhanna
- Have to purge sensitive info e.g. keys from structures. Common Criteria pays lots of attention to policy... might need to extend in future because of new policy capabilities.
- Will
- Audit... just have some info about what policy is enforced... arguing about storing authorization data as blob, consuming system should log.
- Shawn
- Simo, do you have generic authorization check calss like gss_userok that we could tie authorization checks into?
- Simo
- not using generic interfaces. Also could increase size of logs. A ticket is used multiple times.
- Shawn
- What are concerns about sensitive data?
- Simo
- PAC has groups etc. Type of authentication: smart card etc. might want to audit.
- Shawn
- Different interface for auditing authorization data and stick with proposed design otherwise.
- Shawn
- Requirements for ticket ID, e.g. CAMMAC.
- Tom
- Didn't think that was in actual document. We'd have to think about what it means.
- Shawn
- Hash?
Tom describes differences between using a ticket-ID per initial authentication event, versus some simple hash that is possibly implementation dependent and needs correlation by postprocessing.
- Shawn
- Prefer internal ID.
- Simo
- If used as authorization data... does it actually bind to a ticket?
Tom will write up more about CAMMAC and types of binding elements. Also will reread MS-PAC.
- Will
- Do we have consensus?
- Tom
- Seems to be design 3.
- Will
- Performance
- Greg
- Need to check with Dmitri
- Zhanna
- If enabled
- Will
- For us, always enabled. We're going to need kadmin auditing.
- Zhanna
- Next step after KDC auditing