Re: [Model-t] draft-thomson-tmi

[Christian Huitema <[e-mail]>]

On 7/22/2020 5:35 AM, Martin Thomson wrote:
> On Wed, Jul 22, 2020, at 11:13, Christian Huitema wrote:
>> Now, voice 
>> codec just send at a constant bit rate, independent of the variations 
>> in speech. 
> I don't believe that this is true.  I don't think that countermeasures against these attacks are very widely implemented.  There is definite leakage, particularly for things like IVRs, so that might be worrying.  Of course, my information is a little dated.
>
>> Unless of course there is a clear requirement that devices can be 
>> audited -- and we need regulations for that.
> That seems to be the only real solution here.

Yes. I see the other messages from Colin and Carsten pointing out that
the audio leakage issue is not always fixed in encrypted voice, despite
being well known. I stand corrected and my trust in VoIP services is
notched down one point lower still, but the general point stands. The
general point is that a lot of the device inspection work going on today
amounts to exploiting bugs in the devices. Bugs in the PKI
implementation allow research to conduct MITM attacks and inspect the
traffic, absence of masking traffic or constant length padding allow for
detection of voice activity. There are obviously many bugs in at least
some devices, which can be used in practice for some inspection. But I
don't believe we can rely on that forever. If privacy researchers can
exploit the bugs, so can a variety of hackers. We would rather get the
bugs fixed. But then, we do need a requirement to allow inspection.

My other point is that the most egregious leakage happens in the cloud.
Yes, like Mirja suggests, I would personally not install cloud backed
devices in my house. But that's a trade-off between function and safety,
and today I don't believe we can impose that trade-off to the public at
large. That mean that if there is to be regulation of devices, it has to
address the cloud as well.

-- Christian Huitema