Re: [Anima] creating iDevID certs with openssl

[Robert Moskowitz <[e-mail]>]

On 08/25/2017 09:36 AM, Kent Watsen wrote:
>
>
>
>> On the openssl-user list, the claim was made that concatinating DER
>> certs for a cert chain like you do with PEM does not work with
>> applications.  You have to use PKCS#12, but that includes the private
>> key and is passworded.  I should be able to get the pointer to Viktor's
>> post about this.
> Concatenating PEM encodings into a single file is a hack, albeit a super
> convenient one.  Instead of PKCS#12, have you looked into PKCS#7? - this
> is what is used in the netconf-zerotouch draft for binary certificate
> chains.

Please read:

https://mta.openssl.org/pipermail/openssl-users/2017-August/006374.html

VIktor gives his views of what works in the field.  I have found Viktor 
to be a great help, and tends to really know what works. Besides on the 
openssl-user list, he is on the postfix-user list.
>
>
>> Here is where you'd find objects in BRSKI:
>>
>> 1) TLS ClientCertificate.
>>      This should be just the IDevID blob, and in TLS, it's in DER format,
>>      if the Registrar needs anything else in the chain, it must chase them down
>>      itself.
> FWIW, the netconf-zerotouch draft specifies that the device possesses the
> IDevID cert *and* all the intermediate certs leading to the manufacturer's
> well-known trust-anchor, all of which it provides during crypto handshake.

Most systems running netconf have the storage capacity for this.  I 
would think.

In my auto meetings, many of the devices they are discussing do not have 
this option.


>
>> 2) TLS ServerCertificate.
>>      https://tools.ietf.org/html/rfc5246#section-7.4.2
>>        Note: PKCS #7 [PKCS7] is not used as the format for the certificate
>>        vector because PKCS #6 [PKCS6] extended certificates are not used.
>>        Also, PKCS #7 defines a SET rather than a SEQUENCE, making the task
>>        of parsing the list more difficult.
> I don't understand how this relates to PKCS#6.  True about the SET, but this
> is minor relative to the tradeoffs.  Can you say some more about this?
>
>
>> 3) The plege's Voucher Request may be signed using JOSE (using the IDevID)
>>      The IDevID is not sent, it's the one from ClientCertificate.
>>
>> 4) The registar's Voucher Request may be signed using JOSE, using the
>>      Domain Owner's key.  That might not be the same key the Registrar uses
>>      to form the EST connection to the MASA (if the Registrar uses client
>>      authentication at all).
>>      The Domain Owner is within the pinned-domain-cert field.
>>      We define it as being DER binary.  In JSON format, that turns into base64
>>      encoded.  The reason we go there is that we do not want the
>>      ----BEGIN... stuff in there for JSON, in another encoding (CBOR), binary
>>      is just fine.
> Correct, putting PEM-text into protocols is inappropriate.
>
>
>> 5) The resulting voucher also uses pinned-domain-cert as well, now signed
>>      by the MASA.
>>
>>       > What format are the various bootstrap objects (eg vouchers)?  Has
>>       > anyone built any of these for PoC?
>>
>> PoC?
> Aren't the formats defined in the drafts?  E.g., the voucher draft
> says it’s a PKCS#7.

I think PKCS is mentioned twice in the bootstrap draft.  Nothing about 
iDevID format, for example.  802.1AR does not say anything about iDevID 
(or lDevID) format.

> Regarding PoC, here's some code to create/validate vouchers:
> https://github.com/netconf-wg/zero-touch/tree/master/openssl-test.

For next week.

Bob