Re: [Lurk] lurk integration with openssl

Hi Jesus,

That is really great to have two implementations! In addition the c
implementation integrated with openssl would provide more accurate
measurements on how lurk impact the edge server. That is great news and I
wish interoperability tests may be done at the hackathon!

Yours,
Daniel

On Wed, May 23, 2018 at 11:38 AM, Jesús Alberto Polo <ietf@jesusalberto.me>
wrote:

> Hi,
>
> A very early version of clurk will be ready very soon. So far, the ECDHE
> handshake is done (POO is missing though) and I’m finishing the RSA
> handshake (PFS is missing as well) based on the patch you shared lately.
>
> I’ll share the GitHub link as soon as it is ready and also provide more
> details.
>
> Best,
>
> Jesús Alberto
>
>
> On 2018-04-24 17:34, Daniel Migault wrote:
>
>> Thanks for the feed back! Yes absolutely for ecdhe, the sig_and_hash
>> is missing from the spec. I have also slightly changed the extended
>> master structure by exchanging the session_hash and encrypted
>> premaster. I expect to be able to update the draft by next week as
>> well. On my python implementation I am using the following structures
>> for ecdhe.
>>
>> Yours,
>>
>> Daniel
>>
>> TLS12ECDHERequestPayload = Struct(
>>
>>     Embedded(TLS12Base),
>>
>>     "sig_and_hash" / SignatureAndHashAlgorithm,
>>
>>     "ecdhe_params" / ServerECDHParams,
>>
>>     "poo_params" / Struct(
>>
>>         "poo_prf" / Default( POOPRF, "null" ),
>>
>>         "rG" / IfThenElse( this.poo_prf == 'null',
>>
>>              Pass,
>>
>>              Switch( this.ecdhe_params.curve_param.curve,
>>
>>                 {
>>
>>                 "secp256r1" : UncompressedPointRepresentation_256,
>>
>>                 "secp384r1" : UncompressedPointRepresentation_384,
>>
>>                 "secp512r1" : UncompressedPointRepresentation_512
>>
>>                }) ),
>>
>>         "tG" / IfThenElse( this.poo_prf == 'null',
>>
>>               Pass,
>>
>>               Switch( this.ecdhe_params.curve_param.curve,
>>
>>                   {
>>
>>                   "secp256r1" : UncompressedPointRepresentation_256,
>>
>>                   "secp384r1" : UncompressedPointRepresentation_384,
>>
>>                   "secp512r1" : UncompressedPointRepresentation_512
>>
>>                }) ),
>>
>>     )
>>
>> )
>>
>> With
>>
>> TLS12Base = Struct(
>>
>>     "key_id" / KeyPairID ,
>>
>>     "client_random" / Random,
>>
>>     "server_random" / Random,
>>
>>     "tls_version" /  ProtocolVersion,
>>
>>     "prf" / PRFAlgorithm
>>
>> )
>>
>> I have also changed the structure of the extended master by
>> interverting the session hash and the encrypted master to ease the
>> parsing.
>>
>> struct{
>>
>>     KeyPairID key_id
>>
>>     ProtocolVersion tls_version   // see RFC5246 section 6.2.1
>>
>>     PRFAlgorithm prf              // see RFC5246 section 6.1
>>
>>     opaque session_hash<2...2^16-2>
>>
>>     EncryptedPreMasterSecret  pre_master
>>
>>                                   // see RFC5246 section 7.4.7.1
>>
>> }TLS12ExtendedMasterRSARequestPayload;
>>
>>         ]]></artwork>
>>
>> FROM: Jesús Alberto Polo [mailto:ietf@jesusalberto.me]
>> SENT: Tuesday, April 24, 2018 11:11 AM
>> TO: Dmitry Kravkov <dmitryk@qwilt.com>; Daniel Migault
>> <daniel.migault@ericsson.com>
>> CC: LURK BoF <lurk@ietf.org>
>> SUBJECT: Re: [Lurk] lurk integration with openssl
>>
>> Hi,
>>
>> Thanks for the resources and the patch, it’s definitely easier to
>> solve it the way you did in the hackathon.
>>
>> I managed to integrate the basic functionality of LURK for ECDHE and
>> I’m preparing some tests, I hope they’re done and the code cleaned
>> up by the end of this week.
>>
>> Regarding the TLS12ECDHERequestPayload [1], I think the _Signature
>> Algorithm_ field is missing (hash and signature), to indicate the
>> chosen algorithms for the TLS connection.
>>
>> Best regards,
>>
>> Jesús Alberto
>>
>> [1] https://tools.ietf.org/html/draft-mglt-lurk-tls12-00#section-7.1
>> [1]
>>
>> On 22 Apr 2018, 12:08 +0200, Dmitry Kravkov <dmitryk@qwilt.com>,
>> wrote:
>>
>> Hi Jesus Alberto,
>>>
>>> this is a patch for openssl used during 101 hackathon
>>>
>>> It looks that direct calling for lurk library from statemachine will
>>> be hard to push upstream, but adding more callbacks for master
>>> secret calculation that nginx (or other client) registers for,  will
>>> be easier to submit.
>>>
>>> On Fri, Apr 20, 2018 at 9:26 PM Daniel Migault
>>> <daniel.migault@ericsson.com> wrote:
>>>
>>> Hi Jesus Alberto,
>>>>
>>>> There have been some discussions regarding the integration of lurk
>>>> with openssl during the hackathon, so feel free to share your
>>>> concerns on the mailing list..
>>>>
>>>> Here are some links you might find of interest:
>>>>
>>>>
>>>>
>>> https://www.agwa.name/blog/post/protecting_the_openssl_priva
>> te_key_in_a_separate_process
>>
>>>
>>>> https://www.agwa.name/blog/post/titus_isolation_techniques_continued
>>>
>>>>
>>>>
>>>> Yours,
>>>>
>>>> Daniel
>>>>
>>>> _______________________________________________
>>>> Lurk mailing list
>>>> Lurk@ietf.org
>>>> https://www.ietf.org/mailman/listinfo/lurk
>>>>
>>>
>>> --
>>>
>>> DMITRY KRAVKOV
>>> Qwilt | Work: +972-72-2221630 | Mobile: +972-54-4839923
>>>
>>> dmitrykATqwilt.com
>>>
>>
>>
>> Links:
>> ------
>> [1] https://tools.ietf.org/html/draft-mglt-lurk-tls12-00%23section-7.1
>>
>
> _______________________________________________
> Lurk mailing list
> Lurk@ietf.org
> https://www.ietf.org/mailman/listinfo/lurk
>