A prototype for tokenized Treasury products rejected minting, the creation of new tokens, in a synthetic test where a simpler aggregate proof-of-reserves check accepted it. After assets marked as encumbered were excluded, its backing coverage ratio, or BCR, was 96.3408%, below the 105% policy threshold. The aggregate baseline accepted minting in that same scenario.
This was a deterministic scenario, not an estimate from a live portfolio. The evaluation compared RWA-PoB with a simplified aggregate proof-of-reserves baseline using USDY-calibrated liabilities and synthetic reserve scenarios. The reported figure came from that constructed encumbered-assets state.
A rules-based gate on issuance
The controller requires five institutional signers and EIP-712 digital signatures. It applies BCR and Redemption Liquidity Coverage, or RLC, thresholds to issuance and to the choice between immediate and queued redemption.
The design ties token supply to liability accounting. Issuance and the corresponding liability increase are coupled in one transaction. Redemption couples token burning with reclassifying the amount as a pending liability.
The prototype was implemented in Solidity and evaluated with Hardhat on a local Ethereum-compatible network. Its complete deterministic suite contained 31 controller, integration and adversarial, seeded state-machine, and scenario and gas checks, and all 31 passed in post-merge continuous integration.
The tests were built around a calibrated snapshot
The liability calibration used 1,043 daily observations from 18 September 2023 to 26 July 2026. The latest observation was approximately $2.162 billion, and each synthetic scenario contained 100 positions generated with seed 42.
Scenario construction set gross reserves at 115% of observed token value, other liabilities at 0.5%, and the proposed redemption at 5%. The policy thresholds were 105% for BCR and 100% for RLC.
Experiments used the USDY data to calibrate liability scale, but they did not reproduce historical token supply or a changing redemption price. The prototype fixed the redemption price at 1.
Backing and liquidity produced different decisions
In the valid synthetic state, RWA-PoB matched the baseline's decision and permitted minting. Across every evaluated scenario, the aggregate baseline permitted minting, while RWA-PoB rejected it in the encumbered-assets state.
A separate liquidity-stress scenario tested redemption liquidity. Both systems accepted minting, and BCR was 114.1209%. The proposed redemption was queued, with post-request RLC at 39.9999%, below the 100% RLC threshold.
Integration tests showed that submitted redemptions remained pending until settlement. Confirmed payment reduced the outstanding liability and tracked liquid assets. That was the prototype's accounting behavior, not proof of bank finality.
The cost of the extra checks
Gas results showed a mixed trade-off. Initial RWA-PoB publication used approximately 3.83 times the baseline gas, while subsequent updates used about 5.16 to 5.23 times baseline.
In ten reported repetitions, atomic issuance used 132,573 gas, compared with 153,006 gas across two transactions in the comparator. The difference was 20,433 gas, or 13.35%.
These were run-specific local measurements. They excluded deployment, off-chain processing, oracle fees and network prices, so they are not a general fee estimate.
What the checks cannot establish
The central caveat concerns the off-chain records behind each snapshot. Signed snapshots authenticate provenance and policy metadata, but the controller does not independently verify whether off-chain assets exist, are owned as claimed, are complete, correctly valued, have the relevant legal status or have reached bank finality.
Signatures alone cannot establish the truth or completeness of those asset claims. The framework supplies a rule for handling reported information, not independent confirmation of the information itself.
The reported evaluation remains at the prototype-policy level. It covers a comparison with a simplified baseline, synthetic reserve scenarios and local EVM measurements, rather than a live portfolio. It also leaves changing-price redemption outside the experiment. The 31 passing tests do not amount to formal verification or an independent security audit.
The manuscript is an arXiv preprint dated 26 August 2026. The authors state that the prototype, test suite, datasets and replication scripts are available in the [PoB GitHub repository](https://github.com/rischanlab/PoB).
Paper data and sources
Original title: RWA-PoB: A Credential-Based Proof-of-Backing Framework for Tokenized U.S. Treasury Products
Authors: Rischan Mafrur, Gun Gun Febrianza, Sean Foley
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text