Cost of corruption

The cost of corruption of a claim is the cheapest total cost of a set of atom failures that makes its formula false. It is relative to the consumer’s policy: the set of assumptions the consumer is willing to trust.

The cost of one atom

Under a policy, an atom costs:

The standard policy (Policy::standard()) trusts the Cryptographic and Specification classes and nothing else, with κ = 2. The paranoid policy trusts nothing. A consumer can also trust individual atoms, distrust individual atoms (even if their class is trusted), and assume atoms already broken, for example a TEE after a published physical attack, before any adjudicator rules.

Combining

The cost is computed bottom-up:

CoC(atom)    = cost(atom)
CoC(φ1 ∧ φ2) = min(CoC(φ1), CoC(φ2))     breaking any input breaks the claim
CoC(φ1 ∨ φ2) = CoC(φ1) + CoC(φ2)         every alternative must be broken

The sum rule is exact when the alternatives share no priced atom. When they do, the exact value is a minimum-weight hitting-set problem, NP-hard in general. The implementation then reports a sound interval:

• low: the maximum of the alternatives’ costs (breaking φ1 ∨ φ2 means breaking both, which costs at least the dearer one);
• high: the sum.

Every value is reported as { low, high, exact }. The decision rule always uses the low end. sion_assurance::exact_coc computes the exact value by branching over shared atoms when there are few of them. Schemas can require atom-disjoint alternatives, which restores the exact sum.

Two consequences

• Independent evidence adds up. Three disjoint witnesses bonded at 20,000 USD each give a cost of corruption of 60,000 USD. Truth gets harder to fake as participation grows.
• Sybils gain nothing. One party running all three witnesses still forfeits all three bonds once the lie is proven.

What it does not count

Cost of corruption counts stake that is slashable once a falsehood is proven. A lie nobody can disprove costs nothing to tell. Claims that are rarely disprovable carry a detection factor below 1, which the decision rule applies, and agents should treat them as weak.

An attack cost (such as a TEE’s) deters but compensates nobody: only bonds can be slashed.

On the devnet

The receipts demo prices one AI inference receipt, dual-signature ∧ (TEE quote ∨ bond), under three policies:

Pulse  80  receipt issued  [dual-signature ∧ (TEE quote: nvidia-h100 ∨ bond:inference-co)]
           standard policy: cost to forge = 6,000 USD  (TEE attack ≈ 1,000 USD + 5,000 USD bond)
           the TEE's attack cost deters forgery but cannot repay anyone; only the bond is slashable
           standard           relying on 2000 USD → ACCEPT: cost of corruption 6,000 covers κ·V = 4,000
           standard           relying on 3000 USD → REJECT: only 5,000 of backing is unreserved; κ·V = 6,000 is needed
           assumes TEE broken relying on 2000 USD → ACCEPT: cost of corruption 5,000 covers κ·V = 4,000
           assumes TEE broken relying on 3000 USD → REJECT: cost of corruption 5,000 (after detection discount 1000/1000) is below κ·V = 6,000
           trusts the TEE     relying on 2000 USD → ACCEPT: cost of corruption ∞ covers κ·V = 4,000
           trusts the TEE     relying on 3000 USD → ACCEPT: cost of corruption ∞ covers κ·V = 6,000

Under the standard policy, the cost of corruption (6,000 USD) covers κ·V = 6,000 USD in the 3,000 USD case, but the claim is still refused: only the 5,000 USD bond can be reserved, and the TEE’s attack cost cannot (see Reservations).

GET /v1/claims/{id} reports the cost of corruption under the standard policy as coc_standard_policy; POST /v1/claims/{id}/decide takes any policy.