Re: GNUNET_CRYPTO_ecdh_ecdsa fails with anonymous key

[Schanzenbach, Martin]

Hi,


I found the error.
3583e51e3..0b2c0f22f

BR
Martin

> On 2. Mar 2021, at 20:17, Schanzenbach, Martin <mschanzenbach@posteo.de> 
> wrote:
> 
> Hi,
> 
> I can confirm this is happening.
> My first gut feeling also was that this should work.
> The anonymous identity (=private key) is simply "1".
> 
> The public key for the generator G is 1*G = G and also constant.
> Given that ECDH simply multiplies the scalar values, there is not really
> an obvious reason to me why this should not work except for a math thing or
> implementation quirk in libsodium.
> 
> BR
> Martin
> 
>> On 1. Mar 2021, at 17:57, TheJackiMonster <thejackimonster@gmail.com> wrote:
>> 
>> Hi,
>> 
>> I was actually integrating private messages into the messenger API when
>> I encountered that decrypting messages failed awfully. I thought
>> something in my code was wrong, then I double checked the crypto
>> functions:
>> 
>> GNUNET_CRYPTO_ecdh_ecdsa(...) and GNUNET_CRYPTO_ecdsa_ecdh(...)
>> 
>> But their test-case in util checked successfully. So I tested with the
>> ECDSA key from GNUNET_CRYPTO_ecdsa_key_get_anonymous() which gets used
>> by the anonymous ego in the identity service.
>> 
>> So the result was, that GNUNET_CRYPTO_ecdh_ecdsa and
>> GNUNET_CRYPTO_ecdsa_ecdh returned different hashes when using the
>> anonymous key. These functions are using for encrypting and decrypting
>> messages (generating a shared key).
>> 
>> From my point of view it would be fine to just letting the
>> GNUNET_IDENTITY_encrypt() and GNUNET_IDENTITY_decrypt() functions fail
>> when the anonymous key gets used (it doesn't make sense for private
>> messages anyway) but I was wondering why verification still seems to
>> work with this key.
>> 
>> Also I wanted to ask if this behavior was expected. Maybe this could be
>> addressed in some doxygen comments.
>> 
>> Happy hacking
>> Jacki
>

signature.asc
Description: Message signed with OpenPGP