Ironwood programmable shielded experiment: DEA
The strahncryptography account published the Delta Execution Argument (DEA), which hides and proves arbitrary programs in spare rows of a Halo2 circuit. It is still an unofficial experiment.
The author introduces the Delta Execution Argument (DEA) as a construction that breaks the assumption that shielded programmability is expensive. The idea is to define a circuit B that is a functional superset of circuit A, and prove the execution of arbitrary computation on it, so that a single proof is valid for both A and B. The author believes this can be applied by filling roughly 200 padding rows in the Orchard/Ironwood circuit with real computation.
The figures the author gives are 1,152 bytes of extra proof data per Action (a 768-byte delta argument and a 64-byte gate proof). Verification takes 4.7% longer than a normal Action proof, or 11% longer in a batch of 16, and proof generation was about 21 to 26% slower in MockProver tests. Every program yields a proof package of the same size, so the programs cannot be told apart.
Examples tested include spending limits, fee caps, invoices, escrow, deny lists and the Travel Rule. Because the psi field lets a UTXO pass state on to the next spend, the author also wrote a vault program with a per-withdrawal limit, a lifetime limit and a cap on the number of withdrawals. Since it depends on padding rows, the functional limit is about 210 rows, and time-based locks would require putting nLockTime into the Orchard/Ironwood proof.
The post states that it is the operator's personal writing and not an official announcement from Strahn Cryptography. The code is at a rough assembler stage, and the figures come from the author's own tests and have not been externally verified.
This summary is based on the original; check the original for exact details.