maintaining SP shib state in cookie

[Cantor, Scott]

On 5/14/15, 9:40 PM, "Jeremy Shapiro" <jnshapiro at gmail.com> wrote:
>
>We're using DNS load-balancing.

No comment.

>What attack does having the data be tamper-proof and encrypted protect against?   

What do you expect would be in it that wouldn't need to be tamper-proof but yet have no corresponding state on the server to correlate it to? If there's no server state, the cookie *is* the session. It can't be forgeable or there'd be no security in the entire system.

A session in the SP is all of the user's data from the IdP. That also can't be exposed in plaintext to the client. People put things like SSNs into attributes. Heck, they put *passwords* in them.

>> Another option is cross-node lookup of the data to move the state around in memory.
>
> 
>Is this different than the shared process mechanism with private subnets described in the wiki?

I'm talking about implementing lookups between shibd processes to locate non-local data to move it into place. It still requires a private network, or new work to secure a remote call, it just isn't drastically space-limited like a cookie is.

-- Scott