Eve library

Crypto Replacement Targets — V1-P2-0176

Guide

Revision 1

On this page

Audit scope: every site in libs/aje/, libs/aphrodite/, libs/airmid/, libs/iris/memory/privacy/, libs/iris/a2a/, and libs/iris/privacy/security/dlp/ where the source code marks itself as a non-real cryptographic operation: // would use crypto, // would use AES-GCM, // would use keccak256, simulated* helpers, simpleHash, Math.random()- shaped hashes/signatures, and similar placeholder patterns.

Replacement target: every entry below resolves to a primitive in @oshun/crypto (the OshunCryptoSuite introduced by V1-P2-0173, plus the OshunSigningKeystore from V1-P2-0174). Resolution policy:

placeholder pattern replace with
simpleHash(input: string) → number/string keccak256 (eth path) / sha256 (non-eth)
simulatedKeccak256(...) keccak256(...) from @oshun/crypto
simulatedSign(digest, priv) secp256k1Sign(digest, priv) (eth/btc) / keystore
simulatedRecover(digest, sig) secp256k1Recover + keccak256 for address
"would use crypto.subtle.digest('SHA-256')" sha256 from @oshun/crypto
"would use AES-256-GCM" / XOR placeholder aesGcmEncrypt / aesGcmDecrypt
"would use ECDSA / Schnorr to sign" secp256k1Sign / Bitcoin Schnorr (@scure/btc-signer)
Math.random().toString(16) for hash/sig randomBytes(32) for material;
recompute the actual hash from inputs
simulated threshold encryption aesGcmEncrypt + Shamir over secp256k1Ecdh

Each entry below is a discrete checklist item. When a site has been remediated, flip its checkbox to [x] and reference the commit that replaced it.


1. Explicit "would use ..." comments#

Comment-marked placeholders that name the missing primitive directly. These are the highest-priority remediations because the comment is a contract: the production behaviour is not what the code currently does.

1.1 libs/aje/#

  • libs/aje/defi/src/safety/cow-protocol.ts:86 — "Simplified hash representation (actual implementation would use keccak256)". Replace with keccak256(utf8Encode(payload)) from @oshun/crypto and remove the legacy simpleHash/string concat shortcut.
  • libs/aje/wallets/src/social-login/recovery.ts:439 — "XOR encryption (simplified; real implementation would use AES-256-GCM)". Replace with aesGcmEncrypt(key, plaintext, nonce, aad) from @oshun/crypto. Persist the 12-byte nonce alongside the ciphertext; do not reuse across recovery shares.
  • libs/aje/bridges/src/custom-bridge/bridges.ts:215 — "In production, this would use keccak256 and RLP encoding." Replace the message hash derivation with keccak256(rlpEncode(...)). RLP via @scure/rlp if not already available; the keccak primitive is @oshun/crypto.
  • libs/aje/bitcoin/src/bitvm/operators.ts:415 — "Simplified signing (real implementation would use ECDSA)". Replace with secp256k1Sign(messageHash, privateKey) and route private-key access through OshunSigningKeystore.
  • libs/aje/bitcoin/src/bitvm/scripts.ts:434 — "Simplified: real implementation would use secp256k1 point addition". Replace with proper curve point addition (@noble/curves/secp256k1 Point.add).
  • libs/aje/bitcoin/src/lightning/routing.ts:372 — "In a real implementation, this would use crypto.createHash('sha256')". Replace with sha256(buf) from @oshun/crypto.
  • libs/aje/bitcoin/src/lsp/protocol.ts:332 — "In real implementation, would use secp256k1 signing". Replace placeholder with secp256k1Sign(hash, key) (or, when the operator is remote, OshunSigningKeystore.sign).
  • libs/aje/bitcoin/src/lsp/protocol.ts:375 — "In real implementation, would use secp256k1 verification". Replace with secp256k1Verify(sig, hash, pubkey).
  • libs/aje/bitcoin/src/rgb/validation.ts:89 — "In real implementation, would use cryptographic commitment scheme". Replace placeholder commitment with a Pedersen commitment over secp256k1 (secp256k1Ecdh + keccak256 for the binding nonce) or, simpler, keccak256(value || blindingFactor) if hiding-binding is sufficient.
  • libs/aje/bitcoin/src/shared/psbt.ts:188 — "SHA-256 hash (placeholder - in real implementation would use crypto)". Replace with sha256.
  • libs/aje/bitcoin/src/shared/psbt.ts:191 — "In a real implementation, this would use crypto.subtle.digest or a library". Same site — replace the inline implementation entirely with sha256(buf).
  • libs/aje/bitcoin/src/shared/psbt.ts:542 — "In a real implementation, this would use ECDSA or Schnorr to sign". Replace with secp256k1Sign (for ECDSA) or Schnorr via @scure/btc-signer.
  • libs/aje/nodes/src/rpc/alchemy.ts:378 — "In a real implementation this would use crypto, but since we need…". Replace the deterministic pseudo-random fallback with randomBytes for nonces and keccak256 for derived ids.

1.2 libs/airmid/#

  • libs/airmid/clinical/src/audit-compliance.ts:209 — "In production, this would use crypto.subtle.digest('SHA-256', …) but…". Replace with sha256 from @oshun/crypto. Audit-log integrity hashes cannot be simpleHash — clinical compliance contracts assume cryptographic collision resistance.

1.3 libs/iris/memory/privacy/#

  • libs/iris/memory/privacy/src/encryption/zero-knowledge.ts:52 — "Real ZK proofs would use cryptographic circuits." This is correct as a pre-condition comment but the body of the prove function should at minimum sign the public inputs with OshunSigningKeystore so a non-ZK proof is at least attested. Track replacement when a real ZK backend (circom/halo2/risc0) is wired in.

1.4 libs/iris/a2a/#

  • libs/iris/a2a/src/agent-card.ts:352 — "Verify an agent card signature (placeholder - actual implementation would use crypto)". Replace with ed25519Verify (A2A canonical algorithm) over the canonical-JSON-encoded agent card.
  • libs/iris/a2a/src/agent-card.ts:365 — "Check signature presence (actual verification would use cryptographic methods)". Same site — remove the presence-only check and call ed25519Verify.

1.5 libs/aphrodite/, libs/iris/privacy/security/dlp/#

No "would use ..." comments under these directories at audit time (grep returns zero hits). Re-run the scan after large merges before declaring this section complete.


2. simulated* crypto helpers#

libs/aje/sdk/src/core/signer.ts defines a hand-written simulatedKeccak256 and simulatedSign/simulatedRecover. These functions are deterministic — they return reproducible bytes — but they are not collision-resistant keccak and not ECDSA. Every callsite is a stub.

2.1 Helper definitions#

  • libs/aje/sdk/src/core/signer.ts:86 — simulatedKeccak256 definition. Replace the function body with a thin wrapper around keccak256 from @oshun/crypto. After the replacement the function name should be deleted and callers updated.
  • libs/aje/sdk/src/core/signer.ts:197 — simulatedSign. Replace body with secp256k1Sign(digest, privateKeyBytes) (and surface the recovery id from the result). Delete the helper after callers migrate.
  • libs/aje/sdk/src/core/signer.ts:742 — simulatedRecover. Replace with secp256k1Recover(sig, digest) + keccak256(uncompressedPub) → last 20 bytes for the Ethereum address.
  • libs/aje/sdk/src/core/contracts.ts:132 — duplicate simulatedKeccak256. Replace with import from @oshun/crypto.
  • libs/aje/sdk/src/core/events.ts:126 — duplicate simulatedKeccak256. Replace with import.
  • libs/aje/sdk/src/core/factories.ts:130 — duplicate simulatedKeccak256. Replace with import.
  • libs/aje/security/src/mev-protection/encryption.ts:455 — "Encrypt data using a simulated threshold encryption scheme." Replace with a real t-of-n AES-GCM scheme: derive per-share keys via secp256k1Ecdh + hkdf, split via Shamir over the secp256k1 scalar field, encrypt with aesGcmEncrypt. Until a real implementation lands, gate the export so production code paths cannot accidentally rely on it.

2.2 Helper callsites in libs/aje/sdk/src/core/signer.ts#

After 2.1 lands, every call below disappears. They are listed here so the audit can tick them off as the helpers are deleted.

  • libs/aje/sdk/src/core/signer.ts:158 — simulatedKeccak256(privateKeyBytes) (in deriveAddress).
  • libs/aje/sdk/src/core/signer.ts:170 — simulatedKeccak256(combined)
  • libs/aje/sdk/src/core/signer.ts:186 — simulatedKeccak256(pubKey)
  • libs/aje/sdk/src/core/signer.ts:202 — simulatedKeccak256(rInput)
  • libs/aje/sdk/src/core/signer.ts:209 — simulatedKeccak256(sInput)
  • libs/aje/sdk/src/core/signer.ts:265 — simulatedKeccak256(combined)
  • libs/aje/sdk/src/core/signer.ts:324 — simulatedKeccak256(stringToBytes(encoded))
  • libs/aje/sdk/src/core/signer.ts:346 — simulatedKeccak256(stringToBytes(...))
  • libs/aje/sdk/src/core/signer.ts:349 — simulatedKeccak256(bytesVal)
  • libs/aje/sdk/src/core/signer.ts:353 — simulatedKeccak256(nested)
  • libs/aje/sdk/src/core/signer.ts:373 — simulatedKeccak256(stringToBytes(...))
  • libs/aje/sdk/src/core/signer.ts:433 — simulatedKeccak256(encodeDomainSeparator)
  • libs/aje/sdk/src/core/signer.ts:434 — simulatedKeccak256(encodeStructData)
  • libs/aje/sdk/src/core/signer.ts:442 — simulatedKeccak256(combined)
  • libs/aje/sdk/src/core/signer.ts:656 — simulatedKeccak256(combined)
  • libs/aje/sdk/src/core/signer.ts:663 — simulatedKeccak256(reversed)
  • libs/aje/sdk/src/core/signer.ts:712 — simulatedKeccak256(hmacInput) → Replace with hmacSha256(key, message) (the helper is mis-named — it's an HMAC, not keccak).
  • libs/aje/sdk/src/core/signer.ts:718 — simulatedKeccak256(ccInput)
  • libs/aje/sdk/src/core/signer.ts:755 — simulatedKeccak256(combined)
  • libs/aje/sdk/src/core/signer.ts:760 — simulatedKeccak256(reversed)
  • libs/aje/sdk/src/core/signer.ts:766 — simulatedKeccak256(fullPub)
  • libs/aje/sdk/src/core/signer.ts:840 — simulatedSign(digest, …) (legacy tx signing).
  • libs/aje/sdk/src/core/signer.ts:856 — simulatedSign(digest, …) (EIP-1559 signing).
  • libs/aje/sdk/src/core/signer.ts:862 — simulatedKeccak256(unsignedRlp)
  • libs/aje/sdk/src/core/signer.ts:863 — simulatedSign(digest, …) (EIP-2930 signing).
  • libs/aje/sdk/src/core/signer.ts:1108 — simulatedKeccak256(keyIdBytes)
  • libs/aje/sdk/src/core/signer.ts:1115 — simulatedSign(digest, …) (KMS-backed signer fallback). After replacement this call should route through OshunSigningKeystore.sign instead of generating a pseudo-private-key locally.
  • libs/aje/sdk/src/core/signer.ts:1148 — simulatedKeccak256(unsignedRlp)
  • libs/aje/sdk/src/core/signer.ts:1190 — simulatedRecover(digest, sig)

2.3 Helper callsites in libs/aje/sdk/src/core/{contracts,events,factories}.ts#

  • libs/aje/sdk/src/core/contracts.ts:164 — simulatedKeccak256(stringToBytes(str)) (selector / event topic derivation). Real keccak required.
  • libs/aje/sdk/src/core/events.ts:158 — simulatedKeccak256(stringToBytes(str)) (event topic).
  • libs/aje/sdk/src/core/events.ts:163 — simulatedKeccak256(data)
  • libs/aje/sdk/src/core/factories.ts:162 — simulatedKeccak256(stringToBytes(str))

3. simpleHash — non-cryptographic hash used in security paths#

simpleHash is a 32-bit FNV-style or polynomial hash defined separately in many files. It is fine for cache-keying but not acceptable for: tx hashes, addresses, public-key fingerprints, audit log digests, or any value that will be persisted, displayed as 0x…, or compared cryptographically. Each file defining simpleHash is listed first; remediating the definition removes the entire downstream callsite cluster.

3.1 simpleHash definitions#

  • libs/aje/sui-move/src/aptos/aptos.ts:112 — replace with sha256 (Aptos uses SHA-3-256 / SHA-2-256 in its native crypto — adopt keccak256 for tx digests and sha256 for non-tx hashes).
  • libs/aje/sui-move/src/move-language/move-language.ts:84 — replace with sha256 (Move bytecode digests are cryptographic).
  • libs/aje/sui-move/src/sui-network/sui-network.ts:89 — replace with sha256 (Sui uses Blake2b for tx digests; blake3(buf, 32) from @oshun/crypto is the closest acceptable substitute).
  • libs/aje/agents/src/safety/safety.ts — simpleHash(entryData) at L1043. Audit-log entries are subject to tamper-evidence contracts; replace with sha256 and chain via prior-hash field.
  • libs/aje/agents/src/ai/ai.ts — 13 callsites (lines 254, 256, 286, 289, 320, 386, 412, 580, 619, 659, 769, 770, 771, 806, 807, 838, 896, 914, 921, 942, …). All represent tx hashes / model hashes / ZK proof hashes; replace with sha256 (zk-AI proofs typically commit via Poseidon, but sha256 is a strict-better stand-in until a real proving backend is wired up).
  • libs/aje/agents/src/communication/communication.ts — replace with sha256.
  • libs/aje/agents/src/frameworks/frameworks.ts — replace with sha256.
  • libs/aje/agents/src/wallets/wallets.ts — replace with keccak256 (these are EOA address derivations).
  • libs/aje/agents/src/x402/x402.ts — replace with sha256 (HTTP payment header hashes).
  • libs/aje/account-abstraction/src/advanced-aa/advanced-aa.ts — replace with keccak256 (UserOperation hashing).
  • libs/aje/account-abstraction/src/eip7702/eip7702.ts — replace with keccak256 (EIP-7702 authorization tuple hashing).
  • libs/aje/account-abstraction/src/erc6900/erc6900.ts — replace with keccak256.
  • libs/aje/appchains/src/altlayer/altlayer.ts — replace with keccak256.
  • libs/aje/appchains/src/config/config.ts — replace with keccak256 (precompile output digest).
  • libs/aje/appchains/src/da/da.ts — replace with sha256 (data availability commitments).
  • libs/aje/chains/src/abstraction/address-resolver.ts — replace with keccak256.
  • libs/aje/depin/src/iot/iot.ts — replace with sha256 (device attestation digest).
  • libs/aje/gaming/src/mud/mud.ts — replace with keccak256 (MUD Store record key hashing).
  • libs/aje/privacy/src/advanced-privacy/advanced-privacy.ts — replace with keccak256.
  • libs/aje/privacy/src/privacy-pools/privacy-pools.ts — replace with keccak256 (ASP commitment).
  • libs/aje/restaking/src/avs/avs.ts — replace with keccak256 (operator set digest).
  • libs/aje/restaking/src/eigenlayer/eigenlayer.ts — replace with keccak256.
  • libs/aje/security/src/audit-tooling/analysis.ts — replace with sha256 (analysis report digest).
  • libs/aphrodite/payments/src/processors/index.ts:300 — replace private simpleHash with sha256 (PCI-tokenisation surrogate identifiers must be cryptographic).
  • libs/aphrodite/property/src/property-engine.ts:2219 — replace with sha256.
  • libs/aphrodite/safety-intelligence/src/safety-engine.ts:399 — replace with sha256.
  • libs/aphrodite/agent-platform/src/data/security.ts:44 — replace with sha256 (DLP redaction codes; the [HASH:…] placeholder must be a one-way function).
  • libs/airmid/api/src/api-documentation.ts:332 — replace with sha256.
  • libs/airmid/docking/src/qsar-modeling.ts:275 — for QSAR molecular fingerprints simpleHash is acceptable (this is not a security primitive); leave but rename to polynomialHash so future audits don't flag it.

3.2 simpleHash call clusters#

After every definition in §3.1 is replaced by an import from @oshun/crypto, the 325 callsite count drops to zero by construction. The audit can tick these en bloc:

  • libs/aje/sui-move/src/aptos/aptos.ts — 14 callsites at L276, 376, 377, 378, 494, 495, 496, 505, 532, 551, 1890, 1913, 1914, 2316, 2385, 2386.
  • libs/aje/sui-move/src/move-language/move-language.ts — 7 callsites at L191, 455, 456, 461, 477, 494.
  • libs/aje/sui-move/src/sui-network/sui-network.ts — 4 callsites at L478, 1099, 1443, 1529.
  • libs/aje/agents/src/ai/ai.ts — 25+ callsites; verify every txHash/inputHash/outputHash/modelHash/circuitHash/ srsHash/proofData/instanceData/contractAddr is using the cryptographic replacement.
  • All other libs/aje/agents/, libs/aje/account-abstraction/, libs/aje/appchains/, libs/aje/chains/, libs/aje/depin/, libs/aje/gaming/, libs/aje/privacy/, libs/aje/restaking/, libs/aje/security/ files listed in §3.1 — replace every remaining simpleHash( callsite once the definition is gone.
  • libs/aphrodite/payments/, libs/aphrodite/property/, libs/aphrodite/safety-intelligence/, libs/aphrodite/agent-platform/ — same.
  • libs/airmid/api/, libs/airmid/docking/ — same.

4. Math.random()-shaped hashes / signatures / addresses#

These are non-deterministic placeholders that masquerade as cryptographic outputs. They are catastrophic in any code path that treats the result as a commitment, signature, or unique identifier.

  • libs/aje/predictions/src/oracles/oracles.ts:517 — historyHash: '0x' + Math.random().toString(16).slice(2, 66). Replace with '0x' + bytesToHex(keccak256(historyBuffer)).
  • libs/aje/security/src/static-analysis/reporting.ts:1115 — hex generator using Math.random(). Replace with randomBytes(N) then hex-encode.
  • libs/aje/gaming/src/gaming-l2s/gaming-l2s.ts:98 — TxHash mock. Replace with keccak256 over the actual unsigned tx.
  • libs/aje/gaming/src/gaming-l2s/gaming-l2s.ts:101 — Address mock. Replace with keccak256(pubkey).slice(-20) for an EOA.
  • libs/aje/gaming/src/economy/economy.ts:105 — TxHash mock, same fix.
  • libs/aje/gaming/src/assets/assets.ts:108 — TxHash mock, same fix.
  • libs/aje/gaming/src/autonomous/autonomous.ts:147 — TxHash mock, same fix.
  • libs/aje/bitcoin/src/lightning/swaps.ts:437,476 — placeholder signatures and hex strings for Lightning swaps. Replace with secp256k1Sign (signature) and randomBytes for nonces.
  • libs/aje/bitcoin/src/lightning/invoices.ts:542 — invoice payment-hash hex via Math.random(). Replace with randomBytes(32) preimage then sha256(preimage) for the actual payment hash.
  • libs/aje/bitcoin/src/ln-apps/wallets.ts:246,358,368,381 — placeholder signature blobs. Replace with secp256k1Sign over the LN message hash.
  • libs/aje/chains/src/solana/transaction.ts:468,747 — sigPlaceholders = new Uint8Array(numSignatures * 64). These are zeroed signature buffers. Replace with ed25519Sign against the keystore-held key for each signer slot.
  • libs/aje/chains/src/zksync/account-abstraction.ts:347 — "we use a placeholder that represents the bytecode hash". Replace with the actual zkSync bytecode hash routine (keccak256(bytecode) truncated to 32 bytes with the version byte prefix).
  • libs/aje/chains/src/abstraction/unified-provider.ts:497 — "as a placeholder since we don't have the signing key". Route the call through OshunSigningKeystore.sign(keyId, message).
  • libs/aje/bitcoin/src/stacks/wallets.ts:263 — "placeholder that returns a dummy 20-byte hash". Replace with sha256(pubkey).slice(0, 20) for Stacks address derivation.

5. Summary statistics#

directory "would use ..." simulated* simpleHash Math.random mock total
libs/aje/ 13 42 ~280 18 ~353
libs/aphrodite/ 0 0 13 0 13
libs/airmid/ 1 0 7 0 8
libs/iris/memory/privacy/ 1 0 0 0 1
libs/iris/a2a/ 2 0 0 0 2
libs/iris/privacy/security/dlp/ 0 0 0 0 0
total 17 42 ~300 18 ~377

Counts exclude node_modules, *.spec.ts, and *.test.ts. The simpleHash total is reported as ~300 because some callsites are inside template-string expressions where the same line uses the helper multiple times — the authoritative count is the union of the file paths in §3.1, which after remediation will eliminate every occurrence by deletion of the helper.


6. Remediation policy#

  1. Replace the definition first, then the callsites. Because every library has its own simpleHash/simulatedKeccak256, deleting the definition forces every caller in the file to migrate via a single import change. This is faster and less error-prone than touching the call list directly.
  2. Replace eth-shaped hashes with keccak256, others with sha256. Use the table at the top of this document. Anything that lands on the EVM as a hash (tx hash, EIP-712 digest, mapping key, event topic, keccak256(rlp(…))) must be keccak256. Cross-chain or non-EVM digests (Stacks, Aptos, Sui, off-chain audit logs) use sha256 unless the chain spec mandates a different primitive.
  3. Route signing through OshunSigningKeystore. For private-key operations (simulatedSign, "would use ECDSA"), the in-process key was already a placeholder; production code must store keys in a KMS. The replacement is keystore.sign(keyId, message), not a direct secp256k1Sign(message, privateKey) call. Use the noble primitives only in test code that owns the bytes.
  4. Use real randomness. Math.random() for any byte that becomes a commitment, nonce, payment hash, address, or signature is a bug. Replace with randomBytes(N) from @oshun/crypto.
  5. Remove the comment after the fix. A // would use keccak256 comment that survives the replacement is misleading — delete it when the underlying call is migrated.

When every checkbox above is [x], file V1-P2-0176 can be marked complete in OSHUN_V1_TODOS_P2.md. Until then, the backlog of placeholder cryptography is the source of truth and any new code that introduces additional simpleHash/simulated* callers must be rejected at review.