Detect fraud risk

POST

/v1/actions/hospitality.action.detect_fraud_risk/apply

curl --request POST \
  --url https://api.aerandil.com/v1/actions/hospitality.action.detect_fraud_risk/apply \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '{ "branch": "2489E9AD-2EE2-8E00-8EC9-32D5F69181C0", "params": { "band": "low", "confidence": "low", "confidence_score": "example", "coverage_guaranteed": true, "coverage_terms_uri": "example", "expires_at": "2026-04-15T12:00:00Z", "external_record_ref": "2489E9AD-2EE2-8E00-8EC9-32D5F69181C0", "factors_considered": [ "payment_signal" ], "feature_payload_hash": "example", "fire_point": "at_booking", "folio_charge_ref": "2489E9AD-2EE2-8E00-8EC9-32D5F69181C0", "fraud_network_signal_ref": "2489E9AD-2EE2-8E00-8EC9-32D5F69181C0", "person_ref": "2489E9AD-2EE2-8E00-8EC9-32D5F69181C0", "policy_outcomes": [ "psd2_sca_required" ], "prior_triage_ref": "2489E9AD-2EE2-8E00-8EC9-32D5F69181C0", "rationale_text": "example", "reason_codes": "example", "recommended_action": "approve", "reservation_ref": "2489E9AD-2EE2-8E00-8EC9-32D5F69181C0", "risk_context_snapshot": {}, "score": "example", "score_calibrated": "example", "subject_id": "example", "subject_kind": "reservation", "triggered_rules": "example", "vendor_id": "example", "vendor_model_version": "example" }, "return_modified": false, "target": { "id": "2489E9AD-2EE2-8E00-8EC9-32D5F69181C0", "object_type": "hospitality.property" } }'

• Production API

Emit a typed fraud-risk assessment against a reservation (the dominant fire-target), payment, charge, person, loyalty redemption, or a folio with no booking behind it (a group master or house account taking a card hold; the folio must sit at the target property). Targets the property (container); the scoring subject is identified by subject_kind + subject_id params. Inserts a fraud_risk_assessment row in suggested state with an immutable risk_context_snapshot captured at issue time. Validators enforce: subject_kind/subject_id coherence, band/score coherence, recommended_action/band coherence, score and confidence bounds, reason-code contribution bounds, PCI no-PAN in snapshot. Decision lifecycle flows through apply / reject / escalate / expire.

Effects: create hospitality.fraud_risk_assessment; modify hospitality.fraud_risk_assessment; link hospitality.link.fraud_assessment_for_reservation; link hospitality.link.fraud_assessment_for_person; link hospitality.link.fraud_assessment_for_charge; link hospitality.link.fraud_assessment_at_property; link hospitality.link.fraud_assessment_references_external_record; link hospitality.link.fraud_assessment_superseded_by; link hospitality.link.fraud_assessment_for_prior_triage; link hospitality.link.fraud_assessment_consulted_network.

Authorizations

• aerandil_oauth
  actions:apply
• aerandil_oauth
  actions:hospitality.action.detect_fraud_risk:apply

Request Body^required

Media typeapplication/json

object

branch

Apply against an open branch’s overlay view instead of the canonical (main) plane. The action’s writes land in the branch’s copy-on-first-write overlay, its events in the branch’s own stream, and the invocation is journaled for merge-by-replay. Branch writes are action-only, so this is the write-side entry point to a branch (POST /v1/branches forks one).

string format: uuid

params

required

object

band

required

Risk band ordinal

string

Allowed values: low medium high block not_assessed

confidence

required

Operator-facing confidence ordinal

string

Allowed values: low medium high

confidence_score

Calibrated probability [0, 1] (optional)

string format: decimal

coverage_guaranteed

True when the vendor's chargeback-guarantee covers this decision

boolean

coverage_terms_uri

Pointer to the vendor's coverage terms

string

expires_at

Assessment expiry (default per fire-point — see proposal §3.1)

string format: date-time

external_record_ref

Optional pointer to vendor raw payload (external_record_reference)

string format: uuid

factors_considered

required

Signals the assessment conditioned on

Array<string>

Allowed values: payment_signal identity_signal behavioural_signal device_signal velocity_signal geo_signal lead_time loyalty_account_age prior_chargeback_history prior_recovery_history third_party_payer_flag vcn_detected injection_history network_outcome kyc_outcome three_ds_outcome watchlist_match graph_signal fraud_network_signal other

feature_payload_hash

SHA-256 of canonical input-feature JSON (chargeback-defence tamper-evidence)

string

fire_point

required

Lifecycle moment the assessment fired at

string

Allowed values: at_booking at_payment at_check_in at_folio_close on_chargeback_attempt

folio_charge_ref

Optional charge anchor

string format: uuid

fraud_network_signal_ref

The Watchtower answer this assessment was scored with; its band floor binds the band

string format: uuid

person_ref

Optional denormalised person anchor

string format: uuid

policy_outcomes

required

Compliance flags observed during scoring

Array<string>

Allowed values: psd2_sca_required psd2_sca_out_of_scope_moto psd2_sca_exempted_low_value psd2_sca_exempted_trusted_beneficiaries psd2_sca_exempted_tra psd2_sca_exempted_recurring psd2_sca_exempted_corporate_payment three_ds_y three_ds_a three_ds_u three_ds_n three_ds_r three_ds_not_attempted gdpr_art22_explanation_required pci_token_only pci_last4_only facta_red_flag_match fincen_8300_threshold_met eu_ai_act_high_risk_processing ccpa_admt_notice_required none

prior_triage_ref

Optional cross-surface link to a prior message_triage (e.g. injection.suspected)

string format: uuid

rationale_text

required

Human-readable explanation

string

reason_codes

required

Explanation array (JSON array of {code, severity, contribution})

recommended_action

required

Recommended action class

string

Allowed values: approve review step_up_3ds step_up_kyc decline escalate

reservation_ref

Optional denormalised reservation anchor

string format: uuid

risk_context_snapshot

required

Frozen context snapshot at issue time (JSON; host populates per recommended shape)

object

key

additional properties

any

score

required

Raw numeric score on the [0, 100] scale

string format: decimal

score_calibrated

Optional calibrated probability [0, 1]

string format: decimal

subject_id

required

Id of the typed object the assessment scored

string

subject_kind

required

What kind of object the assessment scored

string

Allowed values: reservation payment charge person loyalty_redemption folio

triggered_rules

Custom-rule provenance (JSON array of {rule_id, rule_version, action})

vendor_id

required

Vendor identity (denormalised from Principal::Agent.vendor)

string

vendor_model_version

Vendor model version when exposed

string

return_modified

When true, the response carries modified_objects and modified_links projections — the engine reads the event log inside the same transaction so the projection observes the handler’s own writes.

boolean

target

required

object

id

required

string format: uuid

object_type

required

Allowed value: hospitality.property

Responses

200

Action applied. modified_objects and modified_links populated when return_modified: true was supplied.

Media typeapplication/json

ApplyActionResponse

Envelope returned by the Aerandil action engine. correlation_id ties every event emitted by this invocation; modified_objects and modified_links project the events the handler emitted (per the action’s declared effects); warnings carries any non-blocking soft-validator observations.

object

correlation_id

required

Per-invocation correlation id; matches events.correlation_id for every event this action appended.

string format: uuid

modified_links

required

Links established or removed by this invocation. Empty when the caller did not request return_modified.

Array<object>

object

from_id

required

string format: uuid

from_type

required

string

kind

required

Allowed values: linked

link_type

required

Allowed values: hospitality.link.fraud_assessment_for_reservation hospitality.link.fraud_assessment_for_person hospitality.link.fraud_assessment_for_charge hospitality.link.fraud_assessment_at_property hospitality.link.fraud_assessment_references_external_record hospitality.link.fraud_assessment_superseded_by hospitality.link.fraud_assessment_for_prior_triage hospitality.link.fraud_assessment_consulted_network

to_id

required

string format: uuid

to_type

required

string

modified_objects

required

Objects created, modified, or soft-deleted by this invocation. Empty when the caller did not request return_modified.

Array<object>

object

id

required

string format: uuid

kind

required

Allowed values: created modified

object_type

required

Allowed values: hospitality.fraud_risk_assessment

warnings

Non-blocking soft-validator observations emitted during this invocation (e.g. comp-authority over-limit, recipe pour-cost over sell-price). Absent when no warnings were emitted.

Array<object>

object

message

required

Human-readable summary, suitable for logging or operator-UI surfacing.

string

object_type

required

Object type the warning was raised against.

string

property

Property the warning anchors to, when a single field is the locus. Absent for multi-field rules.

string

rule

required

Stable kebab-case rule identifier (e.g. comp-authority-over-limit).

string

Example

{
  "modified_links": [
    {
      "kind": "linked",
      "link_type": "hospitality.link.fraud_assessment_for_reservation"
    }
  ],
  "modified_objects": [
    {
      "kind": "created",
      "object_type": "hospitality.fraud_risk_assessment"
    }
  ]
}

409

Conflict — concurrent modification or unique constraint violation.

422

Validation error — parameter validation or business-rule violation.