oak_techniques: [OAK-T14.003.001] spec_id: oak-detection-T14.003.001 version: 0.1.0 maturity: emerging maintainer: "@iZonex" license: Apache-2.0 scope: | Detect the LST/LRT depeg-cascade pattern within T14.003 (b), parametrised by which constrained primitive saturates: (i) chain- level redemption-absence (pre-Shapella stETH 2022); (ii) operator- blocked redemption (Renzo ezETH 2024 one-way valve); (iii) withdrawal-queue-depth saturation (Lido stETH / Aave 2025-07). All three sub-classes share the cascade structure secondary-market pressure → bottlenecked redemption → DEX-spot-anchored lending- oracle propagation → looped-leverage liquidation cascade. Excludes: T14.003 sub-cases (a) (c) (d); T9.001 (oracle manipulation — T14.003.001 *uses* DEX-spot-anchored oracles as amplification path but does not require manipulation); legitimate redemption-arbitrage flows that close peg without saturating the constrained primitive. data_sources: [lrt_state, dex_pool_state, lending_market_state, oracle_feed, withdrawal_queue_state, validator_exit_queue, operator_redemption_status] detection_logic: description: | Four orthogonal paths matching the three sub-classes plus the cascade-amplification surface. PATH A (chain-level redemption- absence): the underlying staking primitive does not yet support direct redemption; secondary-market is the only clearing path. PATH B (operator-blocked redemption): deposit-side primitive active while withdrawal-side primitive disabled — one-way valve (Renzo ezETH anti-pattern). PATH C (withdrawal-queue-depth saturation): queue depth × per-period throughput exceeds saturation_threshold of the asset's redeemable supply during the sell-pressure window (Lido / Aave 2025-07 anti-pattern). PATH D (cascade amplification via DEX-spot oracle): lending-market oracle consumes secondary-market price; LTV breaches across looped- leverage positions correlate with secondary-market discount. pseudocode: | # PATH A — chain-level redemption-absence for each LST/LRT L: if not protocol_redemption_enabled(L): sec_market ← secondary_market_volume(L, window = pressure_window) if sec_market > pressure_floor: emit(PATH_A, asset=L, secondary_market_volume_usd=sec_market, severity=high) # PATH B — operator-blocked redemption (one-way valve) for each LRT L: deposit_active ← operator_status(L).deposit_enabled withdraw_active ← operator_status(L).withdraw_enabled if deposit_active and not withdraw_active: sec_pressure ← secondary_market_sell_volume(L, window = pressure_window) emit(PATH_B, lrt=L, deposit_active, withdraw_active, secondary_market_sell_usd=sec_pressure, severity=critical) # PATH C — withdrawal-queue-depth saturation for each LST/LRT L with redemption_enabled(L): queue_depth_eth ← withdrawal_queue_depth(L) throughput_per_day ← per_period_throughput(L) × periods_per_day clear_days ← queue_depth_eth / max(throughput_per_day, 1e-9) sec_discount ← max(0, 1 − dex_price(L) / redemption_rate(L)) if clear_days > saturation_threshold_days and sec_discount > sec_discount_floor: emit(PATH_C, asset=L, queue_depth_eth, clear_days, secondary_market_discount=sec_discount, severity=critical) # PATH D — cascade amplification via DEX-spot oracle for each lending market M with collateral asset L (LST/LRT): oracle_kind ← M.oracle_for(L).kind if oracle_kind ≠ secondary_market_spot: continue sec_discount ← max(0, 1 − dex_price(L) / redemption_rate(L)) ltv_breach ← positions_breaching_ltv_under(M, L, discount = sec_discount) if sec_discount > sec_discount_floor and |ltv_breach| > 0: cascade_usd ← Σ position_size_usd(p) for p in ltv_breach emit(PATH_D, market=M, asset=L, secondary_market_discount=sec_discount, positions_at_risk=|ltv_breach|, cascade_usd, severity=critical) parameters: pressure_window: { type: duration, default: 24h } pressure_floor: { type: number, default: 10000000 } # $10M saturation_threshold_days: { type: number, default: 7 } # > 7 days clear sec_discount_floor: { type: number, default: 0.005 } # 0.5% off redemption output_alert: [oak_technique, detection_path, severity, chain, asset, lrt, market, secondary_market_discount, clear_days, positions_at_risk, cascade_usd, evidence] test_fixtures: positive: - 2022-06-lido-steth-depeg # PATH A — chain-level redemption-absence (pre-Shapella) - 2024-04-renzo-ezeth-depeg # PATH B — operator-blocked redemption (one-way valve) - 2025-07-lido-steth-aave-cascade # PATH C — withdrawal-queue-depth saturation negative: - "Post-Shapella stETH with deep secondary-market liquidity, queue clearing in < 1 day" - "LRT with redemption-rate-aware lending oracle (PATH D suppressed)" - "Healthy LRT with both deposit and withdraw active and < pressure_floor sell volume" false_positive_modes: - PATH A on assets that have intentional permanent redemption-absence as design (e.g., synthetic exposure tokens) — annotate via design_intent - PATH B during scheduled multi-day operator maintenance with deposits paused as well (not one-way valve) — require deposit_active and not withdraw_active simultaneously - PATH C false positives during airdrop / event spikes that resolve without ltv breach — pair with sec_discount > floor as the cascade-firing trigger - DEX-spot-anchored oracles on LSTs whose lending-market exposure is below cascade_margin (PATH D should still emit but at lower severity) mitigations: [OAK-M09, OAK-M11, OAK-M16, OAK-M17] reference_implementations: - { target: gauntlet, chain: ethereum, url: "" } - { target: chaos-labs, chain: ethereum, url: "" } - { target: steakhouse, chain: ethereum, url: "" } - { target: dune, chain: ethereum, url: "" } - { target: lido-monitor, chain: ethereum, url: "" }