#dusk $DUSK I re-read Dusk’s Citadel privacy model documentation and found that the illusion people are most likely to fall for isn’t “how strong the zero-knowledge proofs are,” but rather “who decides which data is visible to whom.”
Citadel lets developers deploy access-controlled smart contracts on $DUSK , using ZK to encapsulate transaction amounts, counterparties, and contract logic within an encrypted boundary. On the surface, this solves the toughest compliance/privacy issues for institutions on-chain—regulators can audit, counterparties can’t see the entire ledger, and the public can verify without needing to know every detail.
But the problem lies exactly in the words “configurable.”
Citadel’s access policies are set by the contract deployer; the privacy boundary isn’t uniformly guaranteed at the protocol level, but instead is pieced together contract by contract and address by address. One contract may expose full data to regulator nodes while only providing hash commitments to regular nodes; another contract might escrow the decryption keys with a third party, with the key-rotation strategy decided by the operator. This is completely different from Tornado Cash’s model of “everyone is invisible to everyone.”
@Dusk takes a selective disclosure approach—privacy is layered, conditional, and revocable. The upside is that institutions can enter in a compliant way; the trade-off is that users must trust that the access policies won’t be modified quietly. The documentation mentions that in the future it may introduce a mechanism like “policy changes require governance voting,” but for now it remains at the roadmap stage.
So I won’t just look at the phrase “zero-knowledge proofs” and conclude that privacy is solved. What needs to be questioned instead is: whether access-policy change records are verifiable on-chain, whether decryption permissions are decentralized, whether the default configuration is conservative enough for ordinary users, and whether an independent audit has validated that Citadel’s permission model can’t be bypassed.
Privacy isn’t a light switch—it’s a permission table. Who can be written into which column is more worth scrutinizing than which encryption algorithm was used.
#dusk @Dusk