Re: [Tcpcrypt] v3 of the charter

[Daniel Kahn Gillmor <[e-mail]>]

On 05/01/2014 04:17 PM, Joe Touch wrote:

> - must not increase information available for fingerprinting: endpoints
> should be identified by their IP address and port and no additional
> information should be provided that indicates the persistence of an
> endpoint or identity other than the IP address and port.

I know i raised the fingerprinting concern earlier here, but i'm
concerned that this statement actually goes too far.  "must not increase
information available for fingerprinting" implies that peers basically
cannot cache state between subsequent connections, which seems like an
obvious opportunity for performance improvements.

When a client connects to a server a second time, it's entirely
reasonable in many cases for the client to presume that the server is
the same server it spoke with last time and be able to prove it by doing
some sort of abbreviated handshake that presumes the server has some
known keying material.  Indeed, the "hooks for authentication" angle
seems to explicitly require the possibility of some sort of persistent
fingerprintability.

How about:

The repeated use of TCP Inc between two peers may provide information to
one peer that could be used to re-identify or "fingerprint" the other
peer in subsequent transactions.  Peers should be able to use TCP Inc
while avoiding leakage of any additional "fingerprinting" information
about themselves if configured or built to do so.

	--dkg