JVET-AL0078 AHG9: On Digital Signing [S. Deshpande (Sharp)]

The following is proposed related to digital signing:

• Proposal 1: Signalling is proposed to allow digitally signing a SEI message.
• Proposal 2: A syntax rearrangement is proposed for syntax elements in the digitally signed content initialization SEI message.
• Proposal 3: It is proposed to signal a tag URI as an identifier in digitally signed content initialization, digitally signed content selection and digitally signed content verification SEI messages. Alternatively, a modification is proposed to the inference of dscs_id.

Additional syntax elements would be added to any SEI message that was desired to be signed. It would be possible to add these syntax elements to newly defined SEI messages. For existing SEI messages, it may be possible to include the additional syntax elements in an extension of the SEI message.

It is proposed to apply proposal 1 to the AI usage restrictions SEI message in VSEI v4.

It is proposed to individually sign each SEI message separately from the rest of the bitstream. It was suggested that it would be possible to remove the SEI message and its signature which would not be able to be detected. A signed SEI message could be added to a completely different bitstream but would still appear valid. It would be unclear what VCL that a signed SEI message applied to.

It was suggested that it would be preferable to associate the SEI message signature with the rest of the bitstream.

Further study is encouraged on proposal 1.

Proposal 2 rearranges some syntax elements to group key related syntax elements. Proposal 2 is agreed for VSEI v4, with the change of moving the substream information to the end of message.

Proposal 3 proposes signalling a tag URI to replace or add to the dscs_id.

The use case proposed for the dscs_id in JVET-AK0206 differs from than the use case described in this contribution. In this contribution, a use case is described for having multiple parties sign.

It was noted that signalling a tag URI would add significant overhead.

It was suggested that an attacker could rewrite the tag URI.

It was suggested that having the ID just indicates and enables that different, independent signing systems could be in use at the same time, while there is sufficient other information in the DSCI SEI messages to identify the different signers. I.e., the ID was not intended to provide sufficient differentiation information for different signers.

The existing semantics of the DSC IDs could be improved to make the intended purpose more clear. This is delegated to the editors. Perhaps can remove “has been produced by a content provider”.

A change is proposed for inference of DSCS verification substream ID, to be equal to the value of dsci_id in the preceding DSCI SEI message, rather than to be equal to 0. JVET-AL0118 also has a related proposal.

It was suggested that if multiple different DSCI SEI messages with different values of DSCI ID are present, the the inference value of DSCS ID will be equal to the DSCI ID of the the DSCI SEI message that is being processed.

Further discussion requested after after side activity led by Y.-K. Wang on inference of DSCS ID.

See the notes on JVET-AL0327.