A one off statement ‘X was not in the database today’ is very different in terms of public confidence and trust to the statement ‘X was not in the database on <<date>>, and no one with access to the database has shown that statement to no longer be true’.
In the field of health privacy ‘verifiable audit’ trails are becoming a buzzword (it’s a blockchain, which…). That goal is derived from the AI world needing an evidence base to convince the public that none of the ‘AI super-intelligence’ they’re trying to build have gone rogue; they use health records as a slightly more palatable justification for it.
Past AI company failures have shown that press releases sometimes don’t match actual events, even fancy cryptographic proofs are not enough to convince critics. There has to be a falsifiable test, which can be easily demonstrated as having passed/failed, in a way which also doesn’t require a detailed understanding of the mechanism itself. In the original ZK-STARKs paper, that would be the equivalent of taking the ‘crooked’ candidate out of the DB, creating the STARK that proved their innocence, and then putting them back in.
That suggests a temporal extension to consideration of ZK-STARKs in the original paper: not only the ability for a STARK to show that the President isn’t in the DNA database; but also a framework to place that STARK in a harness which shows whether someone has produced a conflict over time. Existing similar research seems to avoid the human/corporate nature to undermine a seemingly independent test and the STARK angle…
The obvious analogy is the slashing condition from Proof of Stake used in publicly distributed ledgers: you can make assertions of something, but if you are shown to be wrong, you lose your stake (which may be a peppercorn, but you are seen to lose it. It is a categorical demonstration by the system, even if the details remain opaque. One single honest employee would be able to prove public statements were false.
In a technology world where it is increasingly hard to recruit and retain staff, companies making public statements which such staff know to be untrue will be fundamental disadvantage. All hierarchies fail, generally without the awareness of the people at the top making statements.
The statement ‘X is not in the database’ is very different in terms of public confidence and trustworthiness to the statement ‘X is not in the database and no one has anonymously demonstrated it is false’. However, the glue between the two currently seems underexplored.