The SMPTE FIPS Revisions Study Group was formed several years ago by Tony Wechselberger, DCI’s security consultant. At the time, NIST had just revised its FIPS 140-2 specification, introducing changes of sufficient scope that even DCI stepped out of the shadows and wrote NIST with its concerns. This month, the group concluded its work, issuing an excellent summary that will serve to guide future efforts of similar nature.
Several events took place that greatly helped the group with its work, underscoring the power of organizing such efforts through SMPTE. First, it received expert assistance from around the world. In particular, Mr. Taehyun Kim of DRMinside, South Korean maker of DCP and KDM creation tools, provided expert analysis of every SMPTE standard possibly affected by the changes. Taehyun’s contribution became a significant resource as the group navigated its way forward. Second, while studio execs blustered by making public statements that they had it all under control, Tony knew they were stumped and had the wisdom to reach out to the community for solutions. One of the more potentially damaging of changes was a new requirement that would no longer allow a single digital certificate to be used to decrypt content and enable link encryption. This potentially challenged the very structure of the digital cinema Key Delivery Message (KDM), at worst requiring two different types of KDMs to be used in the field. Fortunately, a clever solution was proposed by Bill Elswick of Entertainment Technology Associates, which would not disrupt the structure of the KDM. Bill’s solution was incorporated in an “errata” of the DCI specification.
A problem of equal significance existed in the intent to obsolete SHA-1 digital signature hash algorithm. FIPS experts stepped in to provide guidance for interpreting the intent of NIST and the actual applications of SHA-1 in digital cinema, eventually leading to the revelation that there would be no impact at this time.
NIST’s last impact to be dealt with was that of random number generation (RNG). Digital cinema encryption allows an optional message integrity check (MIC), in which a key generation process is employed during the decryption process. The key generation process must duplicate that at the point of encryption, thus identical RNGs must be used, with the decryption RNG seeded by a secret value supplied with the encrypted content. A MIC algorithm is then applied using the keys generated by the RNG, from which it can be determined if the encrypted data is intact. To pass FIPS testing, the RNG must be FIPS-approved. The problem is that the RNG must change in new media blocks approved after 2015, causing a MIC failure when the legacy RNG is used in the encryption process.
Fortunately, the change of RNG is a manageable problem. FIPS will allow the legacy RNG to continue to be supported, but some means must be provided in the content to indicate which RNG is used, requiring a change in the standards. But as it is only new media blocks (post 2015) that must incorporate the change, and since the encryption software itself is not FIPS-approved, it’s unlikely that anything but the legacy RNG will ever be used. As it stands today, two standards require updating to accommodate the new RNG: SMPTE ST0429-6 DCP MXF Essence Track File Encryption, and ST0430-5 D-Cinema Packaging – Security Log Event Class and Constraints. The Study Group proposes that no work on changes occur for another year, as there is no rush and it is possible that more changes may be needed.
The great irony of the several changes being made in response to those in the FIPS specification is that none of them will matter to anyone but the bureaucrats. NIST may periodically change the FIPS specification to improve security in the face of ever-increasing computational power. But in practice, the solutions to NIST’s changes that are most acceptable in digital cinema are those that preserve the status quo. It’s our nature. The dual key accommodation in the media block, other than impacting real product designs, will not be monitored as it requires studios to collect security logs, and no exhibitor is sharing security logs today. In addition, the new RNG may impact the design of real products, but, in practice, it too will be ignored in favor of the legacy RNG. This entire exercise has uncovered the real security policy for digital cinema: “do as we say, but watch what we do.”