Re: [rtcweb] draft-ietf-rtcweb-rtp-usage-12 Client-to-Mixer Audio Level

["Cullen Jennings (fluffy)" <[e-mail]>]

On Mar 6, 2014, at 1:57 PM, Magnus Westerlund <[e-mail]> wrote:

> Hi,
> 
> I am not okay with changing 1). I think we MUST mandate implementation
> of Header Extension encryption if you implement the audio level header
> extensions.
> 
> When it comes to use level I am fine with reducing this to RECOMMENDED,
> with a reasonable motivation, like enabling middleboxes that can avoid
> to be in the security context to read these headers. Thus resulting in a
> trade-off between what information one leaks, versus the impact of
> having the middlebox in the security context. It may also need to be
> made clear that the trade-off for this may further be changed as we
> learn more about the feasibility of performing the attack of learning
> the content of the voice communication based on the audio levels.
> 
> I realize that the above is mostly useful if one actually gets a control
> knob for this. But, that is a question for the API.
> 
> From my perspective, allowing for exceptions to the default encryption
> on, when so indicated, is the best way forward.
> 
> Cheers
> 
> Magnus

I agree with Magnus - must implement is fine with me, but because using this feature can have really bad privacy implications, I don’t want this as MUST use. I want the application to be able choose if it uses the feature or not. And of course, the spec needs to have sufficient guidance on the risks.