W3C home > Mailing lists > Public > www-tag@w3.org > December 2014

Re: Draft finding - "Transitioning the Web to HTTPS"

From: Tim Berners-Lee <timbl@w3.org>
Date: Fri, 19 Dec 2014 22:34:29 -0500
Cc: Marc Fawzi <marc.fawzi@gmail.com>, "Eric J. Bowman" <eric@bisonsystems.net>, Chris Palmer <palmer@google.com>, Melvin Carvalho <melvincarvalho@gmail.com>, Mark Nottingham <mnot@mnot.net>, Public TAG List <www-tag@w3.org>
Message-Id: <E7EDA387-9011-4C1B-9536-B524DFFDD2AE@w3.org>
To: Domenic Denicola <d@domenic.me>

On 2014-12 -10, at 12:18, Domenic Denicola <d@domenic.me> wrote:

> From: Marc Fawzi [mailto:marc.fawzi@gmail.com] 
>> - Why does Web Crypto in Chrome depend on https? Transmitting the user's public key over http is how public keys are supposed to be used, in the open. I don't think anyone in their right mind would want to transmit the user's private key (if that's even technically possible... have yet to read about the extractable property and how that works)
> It's not about transmitting the key. It's about transmitting the code that does encryption or decryption. If I can modify that code, I can intercept any supposedly "encrypted" data, or any data that was supposedly meant to be decrypted only on the user's local computer and not sent elsewhere.

Yes, but once the webcrypto code is unpolyfilled into the browser that attack will go away, and you will be able to use
it to build new trust systems, right? 

Webcrypto could completely re-implement the whole stack as it, but also could make different choices.

You could rebuild the CA system just as it is a copy of the current
You could build your own trust tree completely different roots
You could use device pairing protocols
You could build a PGP-like system 
You could build a reputation-based social system
and so on

or, perhaps more importantly, you could build systems which mix any of the above and be able to find an appropriate  way of convincing myself that I should trust, say my bank, a store, a github developer, a friend, a family member, a thermostat, and so on.  The CA system is really at the moment only good for the store.  For my bank I would like to install something special on my machine using a physical device I was given at the physical bank.    For other banks, I'll accept PGP-like recommendation by my bank.  But not by any other CA.     My social network software might allow me to endorse my friends identity.  And so on.   Different models for different cases.  The way these things are bootstrapped must be flexible.


director hat off

Received on Saturday, 20 December 2014 03:34:39 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:33:27 UTC