Re: [Anima-bootstrap] [Spasm] SHA1 usage in Anima-bootstrap voucher yang

[Russ Housley <[e-mail]>]

> On Mar 1, 2017, at 8:52 PM, Michael Richardson <[e-mail]> wrote:
> 
> 
> In the ANIMA ownership voucher YANG model, the absolute latest which you can
> currently see at:
>   https://github.com/anima-wg/voucher/blob/master/draft-ietf-anima-voucher-01.txt
> 
> we write at line 574:
> 
> leaf subject-hash {
>    type binary;
>    description "The certificate's entire subject field MUST
>              match this value.  This value is calculated as the SHA-1
>              hash over the TBSCertificate's subject structure, as
>              specified by RFC 5280 Section 4.1.2.6, encoded using
>              the ASN.1 distinguished encoding rules (DER), as
>              specified in ITU-T X.690.
> 
>              Note: by using the SHA-1 algorithm, the result can be
>              easily compared to OpenSSL's 'subject_hash'
>              output.";
> }
> 
> The voucher is a signed artifact (PKCS7? JWT? CWT? TBD) which indicates to a
> particular device who the devices owner is. ("Are you my mummy?" for Dr.Who Fans)
> 
> For reasons of key hygiene and longevity, in many cases we do not want to
> point at the public key of the registrar directly, but rather indicate that
> it's DN=Foo, as signed by CA=Bar.   It seems like there ought be a better way
> to do this kind of thing than what we specify above, which is annoyingly SHA1
> linked.
> 
> Can SPASM offer any advice?
> 
> We could list the actual DER itself. Encoded, it might actually not be bigger
> than a SHA256, for instance.  That might have privacy implications though,
> which we'd need to think through.
> 
> Some time ago, I proposed replicating the SIDR artifact (RFC3779), and copy
> and pasted it to make:
>    https://www.ietf.org/archive/id/draft-richardson-anima-idevid-cert-00.txt
> 
> but, we didn't really want to go exactly that way.

Michael:

I’m sure you know that there are three important properties for hash
functions.  The are:

(1) collision resistance: it is computationally infeasible to find two
    different inputs that hash to the same output; that is, it is really
    hard to find a and b such that H(a) = H(b).

(2) preimage resistance: it is computationally infeasible to find any
    input that hashes to a given output; that is, given y, it is really
    hard to find x such that H(x) = y.

(3) second-preimage resistance: it is computationally infeasible to find
    a second input which has the same output as a specified input; that
    is, given x, it is really hard to find y such that H(x) = H(y).

Google has announced a collision for SHA-1.  They found to PDF files
that produce the same SHA-1 hash value.

In the system you describe, it seems that an attacker would need to
find a preimage.  For SHA-1, we do not know of a way to do that yet,
but the 160-bit have value produced by SHA-1 is probably not big enough
to be considered safe in today's computing environment.

It seems very odd to be developing a new standards that is using a hash
function that was deprecated at the end of 2010 by NIST.

My personal recommendation ould be to move from SHA-1 to SHA-256.

Russ