Re: [rtcweb] Consensus call regarding media security

["Fabio Pietrosanti (naif)" <[e-mail]>]

On 3/29/12 2:35 PM, Magnus Westerlund wrote:
> On 2012-03-29 08:02, Bernard Aboba wrote:
>> I agree with proposition #1 (SRTP) unconditionally.
>>
>> With respect to proposition #2 (DTLS-SRTP), perhaps the words "with
>> details to be worked out" should have been added. 
>>
>> I believe that the consensus achieved was only on a general direction,
>> not an endorsement of particular proposals.
>>
>> Personally, I would like to have more specifics about the required
>> features of DTLS-SRTP in the RTCWEB context.
> 
> I hope someone that knows the details can elaborate on this. I thought
> DTLS-SRTP has a core that you will need to implement. Then there is
> clearly a question of crypto algorithms to be supported. But that also
> applies to SRTP where we also need to select which crypto suites that
> are to be implemented if any in addtion to the MITM. The WG will need to
> select these details as part of the next steps.

Hi Magnus,

imho the selection of a crypto algorithms is not the first priority (it
will probably goes with some well known AES-stuff), as the first
priority is to define which will be the Key Management system to be used
for DTLS-SRTP.

Given the SRTP standard,the security and trust model radically change
depending on the key exchange system details.

With DTLS-SRTP the key exchange and security/trust model seems quite
opaque and there's not a specific guideline to implement it and/or to
satisfy a specific security requirement.

While the SDES-SRTP method satisfy  all the 4.1.1.* Calling method
requirement of
http://tools.ietf.org/html/draft-ietf-rtcweb-security-02#section-4.1.1 ,
because it would just rely on existing TLS/HTTPS method.

So it would not even enter into the context/discussion of trust with the
server, that has been already managed by the existing TLS/HTTPS.

I did not really understand how DTLS-SRTP would like to manage the
so-many-complicated trust scenario that TLS/HTTPS already manage.

Imho everything should be simplified with two context:
- Do you need end-to-site (TLS equivalent) security?
  SDES-SRTP + TLS/HTTPS

- Do you need end-to-end (ZRTP/PGP equivalent) security?
  DTLS-SRTP with user-driven verification schema (such as Short
Authentication String / Fingerprint / Key pinning)

-- 
Fabio Pietrosanti
Founder, CTO

Tel: [phone] (direct)
Mobile: [phone]
E-mail: [e-mail]
Skype: fpietrosanti
Linkedin: http://linkedin.com/in/secret

PrivateWave Italia S.p.A.
Via Gaetano Giardino 1 - 20123 Milano - Italy
www.privatewave.com