Re: [rtcweb] Tunnelling DTLS in SDP

[Roman Shpount <[e-mail]>]

On Mon, Apr 4, 2016 at 12:14 PM, pfh <[e-mail]> wrote:

>
> On 4 Apr 2016, at 16:52, Roman Shpount <[e-mail]> wrote:
>
>
> I think sending certificate in the answer instead of data path does not
> compromise security since if signaling path is compromised, data traffic
> can be redirected to MITM agent which can collect exactly the same data.
>
>
> That isn’t quite the same, currently in a MITM’d scenario you have the
> option of not completing the DTLS handshake (assuming you had a way
> to verify the fingerprints.). With this change you’ll have all the certs
> logged in a web server somewhere.
>

In the scenario I had in mind MTIM will not modify fingerprints, it will
modify the transport addresses or put itself on the media path using a
compromised router so that it can log the DTLS handshake (and subsequent
media). This will provide exactly the same security risk as logging
signaling exchange when certificate is sent in the signaling path. In order
to compromise media, attacker needs access to the media path. Once it gets
access to the media path, it gets access to DTLS handshake and
certificates. If one of those certificates is sent over signaling channel,
it does not make the whole flow any less secure. Actually you might argue
it makes it more secure, since now attacker needs to compromise both media
and signaling paths in order to decode.
_____________
Roman Shpount