OAK — OnChain Attack Knowledge

Document

OAK v0.1 — Coverage Matrix

This document publishes the honest coverage status of every Technique in v0.1, mapped against the first reference implementation (mg-detectors-rs). Coverage is reported as one of:

  • full — implementation covers the Technique end-to-end, with calibrated thresholds and tests.
  • partial — implementation covers the Technique under restricted conditions or with un-calibrated thresholds; gaps documented.
  • gap — Technique is documented in OAK but no production reference implementation exists yet; calibration and detection methodology are open work items.

OAK's credibility depends on reporting these honestly. Vendors who submit Reference-implementation entries are expected to follow the same convention.

mg-detectors-rs is an operator-behaviour detector; its scope cleanly covers T1–T8. T9 (Smart-Contract Exploit) is a protocol-layer Tactic and is uniformly gap against MG at v0.1 — protocol-layer detection is provided by other tooling (Forta, OpenZeppelin Defender, BlockSec PhalconHQ, audit firms with continuous-monitoring services). T10 (Bridge), T11 (Custody), T12 (NFT), T13 (Account Abstraction), and T14 (Validator/Staking) are similarly out of MG's operator-behaviour-detector scope and are uniformly gap for v0.1 against MG; vendor-side coverage from protocol-and-infrastructure-monitoring tools is the v0.x evolution surface. The two-dimensional matrix planned for v0.5 will surface this orthogonality cleanly.

This document covers the Techniques axis specifically. The other v0.1 axes — Mitigations (40 entries), Software (40 entries), Threat Actors (18 entries), Data Sources (12 entries), and Worked Examples (142 entries) — are not "covered" in the implementation-status sense; they are tracked as catalogue entries with citation-quality verification status. See the Other-axes section below for the per-axis status convention.


Tactic-level summary (v0.1)

Tactic Techniques Full Partial Gap
OAK-T1 — Token Genesis 5 2 3 0
OAK-T2 — Liquidity Establishment 4 1 1 2
OAK-T3 — Holder Capture 3 3 0 0
OAK-T4 — Access Acquisition 6 1 1 4
OAK-T5 — Value Extraction 6 3 0 3
OAK-T6 — Defense Evasion 4 0 0 4
OAK-T7 — Laundering 6 0 1 5
OAK-T8 — Operational Reuse 2 1 0 1
OAK-T9 — Smart-Contract Exploit 5 0 0 5
OAK-T10 — Bridge and Cross-Chain 5 0 0 5
OAK-T11 — Custody and Signing Infrastructure 3 0 0 3
OAK-T12 — NFT-Specific Patterns 3 0 0 3
OAK-T13 — Account Abstraction Attacks 3 0 0 3
OAK-T14 — Validator/Staking/Restaking Attacks 3 0 0 3
Total 58 11 6 41

T6 (Defense Evasion) is intentionally not populated with standalone Techniques in v0.1; evasion patterns are listed as parent tactics on the Technique they modify (e.g., OAK-T2.002 lists T2 + T6 as parents). Standalone T6 entries are scheduled for a v0.x update.

T9 (Smart-Contract Exploit) and T10 (Bridge and Cross-Chain) are uniformly gap against mg-detectors-rs because protocol-layer and bridge-infrastructure detection are out of MG's scope. The expected v0.x evolution is to surface vendor-side coverage from protocol-monitoring tools (Forta, OpenZeppelin Defender, BlockSec PhalconHQ) and bridge-specialist audit firms (Halborn, Trail of Bits, OpenZeppelin) via the multi-implementation matrix planned for v0.5.


Per-Technique status

Technique Status Reference implementation Notes
OAK-T1.001 — Modifiable Tax Function full mg-detectors-rs D01 Static analysis + simulation; calibrated.
OAK-T1.002 — Token-2022 Permanent Delegate full mg-detectors-rs D01 + D07 Binary check on mint extension list.
OAK-T1.003 — Renounced-But-Not-Really (Proxy-Upgrade Backdoor) partial mg-detectors-rs D03 (EVM) Covers: EIP-1967 admin-slot direct read, transparent-proxy / UUPS authority enumeration, OpenZeppelin Ownable-style multi-role mapping on EVM. Excludes: non-EIP-1967 custom proxy patterns, Solana program upgrade authority (different model), diamond-pattern (EIP-2535) facet authority. Token Sniffer / GoPlus / RugCheck cover the EVM standard-proxy surface.
OAK-T1.004 — Blacklist / Pausable Transfer Weaponization partial mg-detectors-rs D04 Covers: ERC-20 static transfer-gate analysis (blacklist mapping reads, paused() guards) + dynamic transfer simulation against deployer / random / victim address tuples on EVM. Excludes: ERC-721/1155 transfer-gate variants, Solana SPL freeze-authority (covered separately under T1.002 family), upgrade-introduced blacklist (post-deploy-only via T1.003). Token Sniffer / GoPlus / RugCheck cover the ERC-20 surface.
OAK-T1.005 — Hidden Fee-on-Transfer partial mg-detectors-rs D04 Covers: ERC-20 static fee-branch predicate enumeration + multi-axis (caller / counterparty / size / time) simulation matrix on EVM. Excludes: SPL Token-2022 transfer-fee extension (covered under T1.002), ERC-721/1155 (no canonical fee-on-transfer surface), rebasing-token mechanic (separate failure mode). Token Sniffer / GoPlus / RugCheck / Honeypot.is cover the EVM ERC-20 surface.
OAK-T2.001 — Single-Sided Liquidity Plant full mg-detectors-rs D10 Threshold calibration per chain documented.
OAK-T2.002 — Locked-Liquidity Spoof partial mg-detectors-rs D02 Cross-chain locker references not yet covered.
OAK-T2.003 — Cross-Chain Locked-Liquidity Spoof gap out of MG single-chain scope Cross-chain lock-receipt-to-pool resolution; unified cross-chain locker registry with per-chain bytecode hashes does not exist as of v0.1.
OAK-T2.004 — Initial-Liquidity Backdoor gap out of MG production scope Pool-creation-transaction trace + LP-mint-recipient + per-admin-role enumerator are documented detection primitives but not production in MG; Token Sniffer / GoPlus / RugCheck cover the EVM Uniswap-V2 / V3 pool-creation surface only, no Solana / Sui / Aptos pool-creation parity.
OAK-T3.001 — Sybil-Bundled Launch full mg-detectors-rs D08 Funder-graph depth calibrated for EVM + Solana.
OAK-T3.002 — Wash-Trade Volume Inflation full mg-detectors-rs D05 Single-venue; multi-venue gap documented in Technique page.
OAK-T3.003 — Coordinated Pump-and-Dump full mg-detectors-rs D04 Confidence-band reporting per Technique page.
OAK-T4.001 — Permit2 Authority Misuse full mg-detectors-rs D12 On-chain layer; wallet-UX layer out of scope.
OAK-T4.002 — Compromised Front-End Permit partial mg-detectors-rs partial Requires off-chain DNS/hosting telemetry input.
OAK-T4.003 — Address Poisoning gap wallet-UX layer, out of MG scope Detection lives at the wallet UX (full-address verification, lookalike detection); on-chain detection is post-event only.
OAK-T4.004 — Allowance / Approve-Pattern Drainer gap wallet-UX layer, out of MG scope Approval-event monitoring + per-spender-cluster heatmaps live at wallet vendors and risk-team layer.
OAK-T4.005 — setApprovalForAll NFT Drainer gap wallet-UX layer, out of MG scope Marketplace-operator allowlisting is the canonical wallet-side mitigation; Discord-compromise vector requires off-chain integration.
OAK-T4.006 — WalletConnect Session Hijack gap wallet-UX + app-store-moderation layer, out of MG scope Per-dApp domain / app-store-package allowlisting at session-establishment time; chains into T4.001 / T4.004 / T4.005 for the actual extraction.
OAK-T5.001 — Hard LP Drain full mg-detectors-rs D02 Real-time + pre-event correlation with T2 signals.
OAK-T5.002 — Slow LP Trickle Removal gap calibration in progress Window/threshold calibration is open research item.
OAK-T5.003 — Hidden-Mint Dilution full mg-detectors-rs D06 Per-chain mint-authority semantics covered.
OAK-T5.004 — Sandwich / MEV Extraction full mg-detectors-rs D13 Public datasets; well-characterized in literature.
OAK-T5.005 — Treasury-Management Exit gap calibration in progress Treasury-balance + destination-clustering primitives in MG scope; off-chain budget-document ingestion + stated-purpose reconciliation are open work items.
OAK-T5.006 — Vesting Cliff Dump gap calibration in progress Per-recipient-cluster outflow monitoring around scheduled cliffs; token-unlock-tracker calendar integration (TokenUnlocks/Tokenomist, CryptoRank, DefiLlama Unlocks).
OAK-T6.001 — Source-Verification Mismatch gap out of MG production scope Deterministic-recompile-against-verified-source is the canonical detection methodology and is in scope for MG's static-analysis surface conceptually but not implemented at v0.1 (no production pipeline binding solc / vyper / cairo / anchor toolchain versions to verified-source metadata). Etherscan / Sourcify / Routescan provide the verified-source substrate; recompile-and-compare automation is an open work item.
OAK-T6.002 — Fake Audit-Claim gap off-chain audit-registry verification, out of MG scope Audit-firm public-registry verification is the canonical user-side mitigation.
OAK-T6.003 — Audit-of-Different-Bytecode-Version gap off-chain audit-attestation verification, out of MG scope Audit-firm-published bytecode-hash vs deployed bytecode comparison; machine-readable cryptographic audit attestations as canonical mitigation.
OAK-T6.004 — Audit-Pending Marketing Claim gap off-chain audit-claim verification, out of MG scope Verifiable audit-engagement attestations + audit-firm public-registry verification + post-listing audit-status decay monitoring as canonical user-side and venue-side mitigation.
OAK-T7.001 — Mixer-Routed Hop partial mg-detectors-rs cluster-graph only Direct privacy-service heuristics not yet implemented.
OAK-T7.002 — CEX Deposit-Address Layering gap exchange-compliance layer, out of MG scope Per-cluster aggregate inflow tracking + threshold-structuring detection on deposit-amount distribution.
OAK-T7.003 — Cross-Chain Bridge Laundering gap cross-chain forensic layer, out of MG scope Cross-chain attribution graphs + per-protocol illicit-cluster inflow metrics; Chainalysis / TRM / Elliptic provide vendor-side coverage.
OAK-T7.004 — NFT Wash-Laundering gap wallet-and-marketplace layer, out of MG scope Per-trade source-of-funds analysis + per-cluster wash-trade-rate metrics at marketplace layer.
OAK-T7.005 — Privacy-Chain Hops gap cross-chain forensic + CEX-compliance layer, out of MG scope Pre-privacy-chain CEX deposit attribution + post-privacy-chain re-emergence inferential heuristics; CEX delisting / Travel Rule policy.
OAK-T7.006 — DeFi Yield-Strategy Laundering gap DeFi-protocol-compliance + cross-protocol forensic layer, out of MG scope Per-position duration + yield-claim-rate + withdrawal-recipient-divergence detection at protocol layer; per-cluster aggregate cross-protocol flow analysis at forensic-provider layer.
OAK-T8.001 — Common-Funder Cluster Reuse full mg-detectors-rs D08 funder-graph Cluster freshness/decay convention pending v0.x.
OAK-T8.002 — Cross-Chain Operator Continuity gap cross-chain forensic layer, out of MG scope Cross-chain attribution-graph providers (Chainalysis, TRM, Elliptic) provide vendor coverage; OAK-G01 attribution depends on it.
OAK-T9.001 — Oracle Price Manipulation gap out of MG scope Protocol-layer; vendor coverage available externally (Forta, OpenZeppelin Defender, BlockSec).
OAK-T9.002 — Flash-Loan-Enabled Exploit gap out of MG scope Protocol-layer; per-transaction flash-loan-pattern detectors are the canonical vendor approach.
OAK-T9.003 — Governance Attack gap out of MG scope Protocol-layer; governance-monitoring agents are the canonical vendor approach.
OAK-T9.004 — Access-Control Misconfiguration gap out of MG scope Protocol-layer; pre-deployment audit + formal verification + deployed-contract monitoring are the canonical defensive stack.
OAK-T9.005 — Reentrancy gap out of MG scope Pre-deployment: checks-effects-interactions + OpenZeppelin ReentrancyGuard + audit + formal verification. Cross-protocol and read-only sub-variants frequently chain with T9.002.
OAK-T10.001 — Validator / Signer Key Compromise gap out of MG scope Bridge-infrastructure layer; off-chain (validator-operator security posture, anti-phishing) + on-chain (per-bridge threshold-signing monitoring) split.
OAK-T10.002 — Message-Verification Bypass gap out of MG scope Bridge-infrastructure layer; pre-deployment audit + post-upgrade smoke-test are the canonical mitigation stack.
OAK-T10.003 — Cross-Chain Replay gap out of MG scope Bridge-infrastructure layer; per-message binding (chain ID + instance address + nonce) is the canonical mitigation.
OAK-T10.004 — Optimistic-Bridge Fraud-Proof Gap gap out of MG scope Architecture-review layer; longer challenge windows + economically-incentivised challenger network + liveness guarantees as canonical mitigation.
OAK-T10.005 — Light-Client Verification Bypass gap out of MG scope Cryptographic-primitive layer; circuit audit + formal verification + trusted-setup MPC ceremonies + multi-prover redundancy as canonical mitigation. Veridise / Trail of Bits / ZK Security as canonical access.
OAK-T11.001 — Third-Party Signing/Custody Vendor Compromise gap out of MG scope Custody-and-signing-infrastructure layer; mitigation is primarily off-chain CTI + procurement controls + out-of-band destination verification.
OAK-T11.002 — Wallet-Software Distribution Compromise gap out of MG scope Wallet-vendor build-pipeline + signed-binary distribution + reproducible builds is the canonical mitigation; user-side hardware-wallet integration as defence-in-depth.
OAK-T11.003 — In-Use Multisig Smart-Contract Manipulation gap out of MG scope Custodial-multisig modification monitoring + change-control delay between proposal and execution is the canonical mitigation.
OAK-T12.001 — NFT Wash-Trade Volume Inflation gap NFT-marketplace layer, out of MG scope Per-collection same-buyer-same-seller patterns + per-cluster wash-trade rate caps; marketplace-side moderation.
OAK-T12.002 — Fake-Mint / Counterfeit Collection gap NFT-marketplace-moderation layer, out of MG scope Canonical-contract-allowlist enforcement + image-perceptual-hash / metadata-similarity scoring; verified-creator badges.
OAK-T12.003 — Royalty Bypass / Marketplace Manipulation gap out of MG scope Marketplace-side royalty enforcement (Operator Filter Registry-style allowlists); EIP-7572 / cross-marketplace policy alignment.

How to read this matrix as a buyer / risk team

A "full" rating means a production implementation exists, has calibrated thresholds, and is tested against known-positive incidents. A "partial" rating means the Technique is detectable under restricted conditions; the per-Technique page documents the conditions and gaps. A "gap" rating means OAK has named the Technique and characterized observable indicators, but production-grade detection is an open work item — published deliberately, so contributors and other vendors can target the gap.

The T9 gap row is not a failure of OAK; it reflects the deliberate orthogonality between operator-behaviour detectors (which MG is) and protocol-layer detectors (which Forta and similar are). Buyers evaluating coverage across the full OAK taxonomy should expect to combine implementations from both classes.

OAK does not score "vendor coverage" by counting Techniques alone. The coverage methodology spec (planned RFC, Y1 Q3) will define an audited methodology for vendor self-attestation against this matrix.


Other implementations

Other vendors are invited to submit Reference-implementation entries via PR (see CONTRIBUTING.md). At v0.1, the matrix is single-implementation. By v0.5 the matrix will be expanded to a two-dimensional matrix (Technique × Implementation).

Other axes — catalogue and verification status

The Mitigations, Software, Threat Actors, Data Sources, and Worked Examples axes are not "covered" in the detection-implementation sense; they are catalogue entries with citation-quality verification status. The convention:

Axis Entries Status convention
Mitigations (OAK-MNN) 40 Each entry maps to ≥1 Technique; coverage status is per-mitigation deployment (vendor-side) rather than per-Technique.
Software (OAK-SNN) 40 Each entry has named CTI-vendor or government-anchor citations; family naming follows the most-cited public-record convention where one exists.
Threat Actors (OAK-Gnn) 18 Attribution-strength language is explicit per-Group: confirmed (government anchor), inferred-strong (industry-forensic anchor), inferred-weak (single-source; OAK does not publish at v0.1).
Data Sources (OAK-DS-NN) 12 Each entry references the canonical telemetry source and OAK Techniques that consume it.
Worked Examples 142 Each example cites the operator post-mortem (where one exists), industry-forensic walkthroughs, and government / court records (where applicable).

Citation verification status (v0.1 pre-launch audit, complete): all 964 entries in citations.bib carry an explicit per-entry status — verified / verified-with-caveat (publicly accessible via standard browser; returns 401-403 to non-browser HTTP clients) / url-not-pinned (canonical URL pending contributor sweep) / url-broken (residual). Government anchors (CISA AAs, OFAC, DOJ, FBI, HHS, foreign-government joint advisories, court records) are verified at submission time. Bulk URL audit completed via tools/verify_citations.py; 0 entries remain in pending verification state.

Update cadence

This file is regenerated at every minor-version bump (v0.x). For in-flight changes between versions, see open PRs labeled coverage:.