Leading 00 byte in ASN.1 DER encoding of unsigned integers with 1st bit set - Re: rfc5280 serialNumber question.

[David von Oheimb]

Just stumbled over this thread of July.

There are still some (cert etc.) viewer implementations, including parts 
of OpenSSL and Mozilla,
that for certain unsigned integer fields (such as in key material or 
cert serial numbers)
output any leading '00' byte as if it was part of the actual byte string 
content,
even when needed just for ASN.1-internal encoding reasons.

Regarding the output of RSA/DSA/DH/EC key material in OpenSSL,
I filed a fix back in January 2019: 
https://github.com/openssl/openssl/pull/8136
This happens to be my oldest OpenSSL PR that is still not merged,
which is just because it could possibly break third-party scripts.

In contrast, as often in my experience, some influential Mozilla folks 
appear to not really understand the issue
and considered my related bug report and fix invalid: 
https://bugzilla.mozilla.org/show_bug.cgi?id=1520923
although I even gave them the - correct - OpenSSL output of an X.509 
cert serial number with 1st bit set as a reference.

     David


On 23.07.23 21:52, Robert Moskowitz wrote:
> While on my flight to IETF, I made 2 certs identical in content other 
> than the serial numbers.  You can clearly see the 0x00 prepended:
>
>
>     0:d=0  hl=4 l= 331 cons: SEQUENCE
>     4:d=1  hl=3 l= 254 cons:  SEQUENCE
>     7:d=2  hl=2 l=   3 cons:   cont [ 0 ]
>     9:d=3  hl=2 l=   1 prim:    INTEGER           :02
>    12:d=2  hl=2 l=   9 prim:   INTEGER :9F07C87244927FBE
>    23:d=2  hl=2 l=   5 cons:   SEQUENCE
>
>     0:d=0  hl=4 l= 330 cons: SEQUENCE
>     4:d=1  hl=3 l= 253 cons:  SEQUENCE
>     7:d=2  hl=2 l=   3 cons:   cont [ 0 ]
>     9:d=3  hl=2 l=   1 prim:    INTEGER           :02
>    12:d=2  hl=2 l=   8 prim:   INTEGER :1BB3EC3B4F2E645C
>    22:d=2  hl=2 l=   5 cons:   SEQUENCE
>
>
> And after spending lots of times reading the appropriate sections of 
> 5280, I realized I was making this harder in my mind than needed.  So 
> the encoding puts 0x00 in front of a type Integer. The decoder just 
> rips it off and makes sure the resultant value is positive.
>
> But it does increase the DER by a byte.  Interestingly in this 
> specific case the PEM was the same size for both.
>
>
> On 7/21/23 14:20, Erwann Abalea wrote:
>> On 8 bits, if the highest bit is set to 1, then the value is between 
>> 0x80 and 0xFF.
>> In your examples, the highest octet of the integer has its highest 
>> bit set (0xFE in the first example, 0xAE in the second), and 
>> therefore the DR encoding added a heading 00, making the real 
>> length 9 octets (l=9 on third column).
>>
>>
>>
>> On Fri, Jul 21, 2023 at 8:00 PM Robert Moskowitz 
>> <rgm at htt-consult.com> wrote:
>>
>>     I looked at a couple of certs.  I might think that if the first
>>     hex is
>>     "F" then the 1st bit is 1, but:
>>
>>
>>
>>          8:d=2  hl=2 l=   3 cons:   cont [ 0 ]
>>         10:d=3  hl=2 l=   1 prim:    INTEGER           :02
>>         13:d=2  hl=2 l=   9 prim:   INTEGER :FE0E6F3753087370
>>
>>          8:d=2  hl=2 l=   3 cons:   cont [ 0 ]
>>         10:d=3  hl=2 l=   1 prim:    INTEGER           :02
>>         13:d=2  hl=2 l=   9 prim:   INTEGER :AEB77AEE2A3CBCD3
>>
>>
>>     I am not seeing a difference in the serialNumber field length.
>>
>>     On 7/21/23 08:58, Robert Moskowitz wrote:
>>     > Per sec 4.1.2.2
>>     >
>>     >    Given the uniqueness requirements above, serial numbers can be
>>     >    expected to contain long integers.  Certificate users MUST
>>     be able to
>>     >    handle serialNumber values up to 20 octets. Conforming CAs
>>     MUST NOT
>>     >    use serialNumber values longer than 20 octets.
>>     >
>>     >
>>     > At some point some years ago it was pointed out here that
>>     serialNumber
>>     > OID encoding preappends 0x00 if the first bit is a 1.
>>     >
>>     > Does this actually make the serialNumber a byte longer?  Or is
>>     this
>>     > only encoding?  Thus IF that first bit is a 1, obviously the
>>     OID value
>>     > is a byte longer.  But when the serialNumber OID is decoded is
>>     this
>>     > longer value returned or the original value?
>>     >
>>     >
>>     > I am girding up to debate an implementation where the CP says
>>     > serialNumber MUST be unique, and their implementation uses a
>>     20-byte
>>     > SN.  I don't think they take care at all about the value of the
>>     1st
>>     > byte.  I doubt in their testing to date they have generated a
>>     SN in
>>     > that range.
>>     >
>>     > So how does the SN with the added byte get decoded?
>>     >
>>     > thanks
>>     >
>>     >
>>
>>
>>
>> -- 
>> Cordialement,
>> Erwann Abalea.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mta.openssl.org/pipermail/openssl-users/attachments/20231107/bbd75335/attachment.htm>