Content Provenance Profile (CPP) Core

1.1. Problem Statement

Digital media authenticity faces several fundamental challenges:¶

• Self-Attestation Weakness: Systems where creators sign their own claims provide no independent verification. A verifier must trust that the creator's claimed timestamp is accurate.¶
• Metadata Stripping: Social media platforms and messaging applications routinely strip embedded metadata, breaking provenance chains that depend on file-level embedding.¶
• Omission Attacks: Systems that prove individual media authenticity cannot detect when unfavorable evidence has been selectively deleted from a collection.¶
• Terminology Confusion: Terms like "verified" mislead users into believing content truthfulness has been established, when only provenance has been recorded.¶
• Analog Hole Attacks: Photographing screens displaying manipulated content can bypass digital provenance systems that only track file-level operations.¶

1.2. Design Goals

CPP addresses these challenges through the following design principles:¶

1. External Timestamp Verification: Timestamps MUST be anchored to independent RFC 3161 Time-Stamp Authorities, enabling third-party verification without trusting the capture device.¶
2. Omission Detection: The Completeness Invariant mechanism enables detection of deleted events within a collection.¶
3. Offline Verification: All data necessary for verification is included in the Evidence Pack, enabling verification without network access.¶
4. Provenance != Truth: CPP proves when and by what device media was captured. It does NOT prove content truthfulness or scene authenticity.¶
5. Hardware-Backed Security: Private keys SHOULD be stored in hardware security modules where available (Secure Enclave, StrongBox, TPM).¶

1.3. Scope

This document specifies:¶

• The CPP event data model¶
• Hash computation and canonicalization rules¶
• Merkle tree construction and proof verification¶
• RFC 3161 TSA anchoring requirements¶
• Verification procedures for implementers¶
• Self-attested signer identity (SignerInfo)¶
• Chain context for partial submission detection¶
• Depth analysis extension for screen detection (OPTIONAL)¶
• Pre-Publish Verification extension (OPTIONAL)¶

This document does NOT specify:¶

• Application user interface requirements¶
• Network protocols for proof distribution¶
• Key management or certificate policies¶
• Content authenticity claims¶

1.4. Changes from draft-vso-cpp-core-01

This revision incorporates the following changes:¶

BREAKING CHANGE — Merkle Tree Domain Separation: This document mandates domain-separated Merkle hashing (Section 4.6.1). LeafHash MUST be computed as SHA256(0x00 || EventHash_bytes) and internal nodes as SHA256(0x01 || Left || Right). Earlier CPP specification documents (v1.0 through v1.4) used non-domain-separated hashing (LeafHash = SHA256(EventHash_bytes), Node = SHA256(Left || Right)). Those constructions are now deprecated. Implementations using the legacy (non-prefixed) construction MUST migrate to the domain-separated construction defined in this document. Test vectors in Appendix B reflect the domain-separated construction and differ from pre-domain-separation outputs.¶

BREAKING CHANGE — AnchorDigest Verification: Verifiers MUST now perform explicit format and digest matching checks on AnchorDigest, MerkleRoot, and TSA messageImprint fields as specified in Section 7.5. These checks were RECOMMENDED in -01 but are now REQUIRED.¶

• Added Section 7.5 with mandatory format checks, TSA binding verification, and PROHIBITED implementation patterns¶
• Added Normative Authority and Legacy Migration guidance to Section 1.6¶
• Strengthened Domain Separation (Section 4.6.1) with MUST/MUST NOT/DEPRECATED language¶
• Added cross-reference from TSA Verification (Section 7.4) to AnchorDigest verification requirements¶
• Clarified test vector applicability for legacy migration (Appendix B)¶
• Corrected author contact email address¶

1.5. Changes from draft-vso-cpp-core-00

The -01 revision incorporated the following changes:¶

• Added SignerInfo for self-attested identity (Section 4.2)¶
• Added DeviceInfo structure for cross-platform support (Section 4.3)¶
• Added CaptureContext for environmental metadata (Section 4.4)¶
• Added Chain Context for partial submission detection (Section 4.8)¶
• Added Depth Analysis Extension for screen detection (Section 8)¶
• Added Pre-Publish Verification Extension (Section 9)¶
• Added C2PA interoperability mappings (Section 10)¶
• Clarified hardware-backed key storage requirements¶
• Added EXPORT and additional event types¶
• Expanded implementation experience with Android and cross-platform validation¶

1.6. Relationship to Other Specifications

CPP defines its own Merkle tree construction that is NOT compatible with Certificate Transparency [RFC6962]. While inspired by similar principles, CPP uses different domain separation prefixes and padding rules optimized for media provenance use cases. Implementations MUST NOT assume RFC 6962 compatibility.¶

CPP is complementary to the C2PA specification [C2PA]. C2PA tracks edit history of content; CPP proves capture provenance with deletion detection. See Section 10 for interoperability mappings.¶

Normative Authority: Where the cryptographic algorithms defined in this document (domain-separated Merkle hashing, AnchorDigest computation, verification procedures) conflict with earlier CPP specification documents (v1.0 through v1.5), this document takes precedence. The CPP specification series published by VSO serves as the design-level reference; this Internet-Draft is the normative interoperability specification for implementations seeking cross-platform compatibility.¶

Legacy Migration: Implementations using non-domain-separated Merkle hashing (LeafHash = SHA256(EventHash_bytes) without the 0x00 prefix) MUST migrate to the domain-separated construction defined in Section 4.6.1. During a transition period, implementations MAY accept both legacy and domain-separated proofs for verification but MUST generate only domain-separated proofs. Implementations SHOULD log a deprecation warning when encountering legacy proofs.¶

3.1. Addressed Threats

3.2. Explicitly Not Addressed

• Content truthfulness (scene staging, deepfakes)¶
• Device compromise before capture¶
• Key extraction from secure hardware¶
• TSA collusion with adversary¶
• Identity verification (SignerInfo is self-attested only)¶

4.1. Events

An Event is the fundamental unit of provenance in CPP. Events are signed records that capture discrete provenance actions.¶

HashSum = EventHash[0] XOR EventHash[1] XOR ... XOR EventHash[n-1]

4.2. SignerInfo

SignerInfo provides self-attested identity claims. It is OPTIONAL and MUST NOT be interpreted as verified identity.¶

4.3. DeviceInfo

DeviceInfo provides metadata about the capture device. It is OPTIONAL but RECOMMENDED for INGEST events.¶

4.4. CaptureContext

CaptureContext provides environmental metadata at capture time. It is OPTIONAL for INGEST events.¶

4.5. Hash Chain

Events form a hash chain through the PrevHash field:¶

Event 1: PrevHash = sha256:0000...0000 (genesis - 64 zeros)
Event 2: PrevHash = EventHash(Event 1)
Event 3: PrevHash = EventHash(Event 2)

Verification of chain integrity:¶

1. For each event after the first, the verifier MUST compute EventHash of the previous event.¶
2. The computed hash MUST match the current event's PrevHash field.¶
3. If any mismatch is detected, verification MUST fail with status CHAIN_INTEGRITY_VIOLATION.¶

4.6. Merkle Tree Structure

CPP defines its own binary Merkle tree construction optimized for media provenance. This construction uses domain separation prefixes to prevent attacks where leaf values could be confused with internal node values.¶

Important: CPP Merkle trees are NOT compatible with RFC 6962 (Certificate Transparency). Implementations MUST use the exact algorithms specified in this section.¶

LeafHash = SHA256(0x00 || EventHash_bytes)

InternalHash = SHA256(0x01 || Left_bytes || Right_bytes)

function computePaddedSize(treeSize):
    if treeSize == 0:
        return 0  // Invalid - TreeSize MUST be >= 1
    paddedSize = 1
    while paddedSize < treeSize:
        paddedSize = paddedSize * 2
    return paddedSize

function buildTree(paddedLeaves):
    levels = [paddedLeaves]
    current = paddedLeaves

    while current.length > 1:
        nextLevel = []
        for i in range(0, current.length, 2):
            left = current[i]
            right = current[i + 1]
            parent = SHA256(0x01 || left || right)
            nextLevel.append(parent)
        levels.append(nextLevel)
        current = nextLevel

    return levels  // levels[0] = leaves, levels[-1] = [root]

4.7. Anchor Structure

The TSA object MUST include:¶

The MessageImprint object MUST include:¶

4.8. Chain Context

Chain Context is OPTIONAL metadata embedded in Evidence Packs to enable partial submission detection. It describes the event's position within its chain without requiring the full chain for initial assessment.¶

Chain Context is informational when embedded in a single proof. Full completeness verification requires the complete chain (Forensic Export). Verifiers SHOULD use Chain Context to alert users when:¶

• TombstoneCount > 0 (events have been deleted)¶
• ActiveEvents < TotalEvents (some events invalidated)¶
• Only a subset of the chain is presented¶

5.1. JSON Canonicalization

Events MUST be canonicalized using [RFC8785] (JSON Canonicalization Scheme) before hashing.¶

The following fields MUST be excluded from canonicalization:¶

• EventHash¶
• Signature¶

All other fields, including SignerInfo if present, MUST be included. Field names in the canonical event object use PascalCase (e.g., EventID, ChainID, PrevHash).¶

5.2. EventHash Computation

function computeEventHash(event):
    eventCopy = copy(event)
    delete eventCopy.EventHash
    delete eventCopy.Signature
    canonical = JCS_canonicalize(eventCopy)  // RFC 8785
    hashBytes = SHA256(canonical)
    return "sha256:" + lowercase_hex(hashBytes)

The resulting EventHash is a 71-character string: the prefix "sha256:" followed by 64 lowercase hexadecimal characters.¶

Note: SignerInfo, DeviceInfo, and CaptureContext are included in the EventHash computation when present. This ensures these fields are tamper-evident and bound to the TSA timestamp.¶

5.3. LeafHash Computation

function computeLeafHash(eventHash):
    hexStr = eventHash.substring(7)  // Remove "sha256:" prefix
    eventHashBytes = hexDecode(hexStr)  // 32 bytes
    prefixedData = [0x00] + eventHashBytes  // 33 bytes
    leafHashBytes = SHA256(prefixedData)
    return "sha256:" + lowercase_hex(leafHashBytes)

5.4. Internal Node Hash Computation

function computeInternalHash(left, right):
    leftBytes = hexDecode(left.substring(7))
    rightBytes = hexDecode(right.substring(7))
    prefixedData = [0x01] + leftBytes + rightBytes  // 65 bytes
    hashBytes = SHA256(prefixedData)
    return "sha256:" + lowercase_hex(hashBytes)

5.5. AnchorDigest Computation

AnchorDigest is the MerkleRoot value WITHOUT the sha256: prefix, represented as 64 lowercase hexadecimal characters.¶

function computeAnchorDigest(merkleRoot):
    return lowercase(merkleRoot.substring(7))

PROHIBITED:¶

• SHA256(merkleRoot) - This would create double hashing¶
• SHA256(stringEncode(merkleRoot)) - This would hash the string representation¶
• Any transformation other than prefix removal¶
• Mixed case output - MUST be lowercase¶

6.1. TSA Request Construction

The messageImprint in TimeStampReq [RFC3161] MUST contain:¶

• hashAlgorithm: SHA-256 (OID 2.16.840.1.101.3.4.2.1)¶
• hashedMessage: AnchorDigest as 32-byte OCTET STRING¶

TimeStampReq ::= SEQUENCE {
   version         INTEGER { v1(1) },
   messageImprint  MessageImprint,
   reqPolicy       OBJECT IDENTIFIER OPTIONAL,
   nonce           INTEGER OPTIONAL,
   certReq         BOOLEAN DEFAULT FALSE,
   extensions      [0] IMPLICIT Extensions OPTIONAL
}

MessageImprint ::= SEQUENCE {
   hashAlgorithm   AlgorithmIdentifier,  -- SHA-256
   hashedMessage   OCTET STRING          -- AnchorDigest (32 bytes)
}

6.2. TSA Response Processing

Upon receiving TimeStampResp, the producer:¶

1. MUST verify the response status is granted (0) or grantedWithMods (1)¶
2. MUST extract the TimeStampToken from the response¶
3. MUST store the complete DER-encoded TimeStampToken¶
4. MUST extract and store the messageImprint from TSTInfo¶
5. MUST extract and store GenTime from TSTInfo¶

6.3. Single-Leaf Tree Rules

When TreeSize equals 1, the following invariants MUST hold:¶

• LeafIndex MUST equal 0¶
• Proof MUST be an empty array¶
• Root MUST equal LeafHash¶
• LeafHash MUST equal SHA256(0x00 || EventHash_bytes)¶

If any of these conditions fail, verification MUST return INVALID.¶

6.4. Multi-Leaf Tree Rules

For TreeSize greater than 1:¶

• LeafIndex MUST be in range [0, TreeSize-1]¶
• PaddedSize = smallest power of 2 >= TreeSize¶
• Proof length MUST NOT exceed log2(PaddedSize)¶
• Sibling hashes are ordered from bottom (leaf level) to top (root level)¶
• Index parity determines pairing order: even=left, odd=right¶
• All internal nodes use SHA256(0x01 || left || right)¶

7.1. Verification Result Codes

7.2. Event Verification

function verifyEvent(event, publicKey):
    // Step 1: Recompute EventHash
    computedHash = computeEventHash(event)
    if computedHash != event.EventHash:
        return INVALID("EventHash mismatch")

    // Step 2: Verify signature
    hashBytes = hexDecode(event.EventHash.substring(7))
    sigBytes = base64Decode(event.Signature)
    if not verifySignature(publicKey, hashBytes, sigBytes):
        return INVALID("Signature verification failed")

    return VALID

7.3. Merkle Proof Verification

function verifyMerkleProof(eventHash, leafIndex, proof,
                           expectedRoot, treeSize):
    // Step 1: Validate inputs
    if treeSize < 1:
        return INVALID("TreeSize must be >= 1")
    if leafIndex < 0 or leafIndex >= treeSize:
        return INVALID("LeafIndex out of range")

    paddedSize = computePaddedSize(treeSize)
    maxProofLength = log2(paddedSize)
    if proof.length > maxProofLength:
        return INVALID("Proof too long")

    // Step 2: Compute leaf hash with domain separation
    currentHash = computeLeafHash(eventHash)

    // Step 3: Handle single-leaf case
    if treeSize == 1:
        if leafIndex != 0:
            return INVALID("LeafIndex must be 0 for single-leaf")
        if proof.length != 0:
            return INVALID("Proof must be empty for single-leaf")
        if lowercase(currentHash) != lowercase(expectedRoot):
            return INVALID("Root != LeafHash for single-leaf")
        return VALID

    // Step 4: Traverse proof from bottom to top
    index = leafIndex
    for siblingHash in proof:
        if index % 2 == 0:
            currentHash = computeInternalHash(currentHash, siblingHash)
        else:
            currentHash = computeInternalHash(siblingHash, currentHash)
        index = floor(index / 2)

    // Step 5: Compare with expected root
    if lowercase(currentHash) != lowercase(expectedRoot):
        return INVALID("Computed root != expected root")

    return VALID

7.4. TSA Verification

TSA verification ensures the timestamp token was legitimately issued by a Time-Stamp Authority and binds the correct digest. Implementations MUST also perform the format and binding checks specified in Section 7.5 prior to or as part of this procedure.¶

function verifyTSAAnchor(eventHash, anchor):
    // Step 1: Verify Merkle structure
    merkle = anchor.Merkle
    result = verifyMerkleProof(
        eventHash, merkle.LeafIndex, merkle.Proof,
        merkle.Root, merkle.TreeSize)
    if result != VALID:
        return result

    // Step 2: Verify LeafHashMethod
    if merkle.LeafHashMethod != "SHA256(0x00||EventHash)":
        return INVALID("Unsupported LeafHashMethod")

    // Step 3: Verify AnchorDigest == MerkleRoot
    expectedDigest = lowercase(merkle.Root.substring(7))
    if lowercase(anchor.AnchorDigest) != expectedDigest:
        return INVALID("AnchorDigest != MerkleRoot")

    // Step 4: Parse TSA Token (RFC 5652 ContentInfo)
    tsaToken = base64Decode(anchor.TSA.Token)
    contentInfo = parseContentInfo(tsaToken)
    signedData = parseSignedData(contentInfo.content)
    tstInfo = parseTSTInfo(signedData.encapContentInfo.eContent)

    // Step 5: Verify hash algorithm is SHA-256
    if tstInfo.messageImprint.hashAlgorithm != SHA256_OID:
        return INVALID("Unsupported TSA hash algorithm")

    // Step 6: Verify messageImprint == AnchorDigest
    tstImprint = lowercase_hex(tstInfo.messageImprint.hashedMessage)
    if tstImprint != lowercase(anchor.AnchorDigest):
        return INVALID("TSA messageImprint != AnchorDigest")

    // Step 7: Verify CMS signature over TSTInfo
    signerInfo = signedData.signerInfos[0]
    signatureValid = verifyCMSSignature(
        signedData.encapContentInfo.eContent,
        signerInfo.signature,
        signerInfo.signatureAlgorithm,
        extractSignerCert(signedData.certificates, signerInfo.sid))
    if not signatureValid:
        return INVALID("TSA signature verification failed")

    // Step 8: Verify certificate chain (SHOULD)
    certValid = verifyCertificateChain(
        signedData.certificates, signerInfo.sid, trustAnchors)

    if certValid:
        return VALID(genTime = tstInfo.genTime)
    else:
        return VALID_WARNING(genTime = tstInfo.genTime,
            warning = "TSA certificate chain could not be verified")

7.5. AnchorDigest Verification Requirements

This section consolidates the mandatory checks for AnchorDigest, MerkleRoot, and TSA messageImprint consistency. These checks are REQUIRED and failure of any single check MUST result in INVALID.¶

Format Requirements:¶

1. MerkleRoot MUST be a 71-character string consisting of the prefix sha256: followed by exactly 64 lowercase hexadecimal characters (regex: ^sha256:[0-9a-f]{64}$). Verifiers MUST reject MerkleRoot values containing uppercase characters, missing prefix, or incorrect length.¶
2. AnchorDigest MUST be exactly 64 lowercase hexadecimal characters (regex: ^[0-9a-f]{64}$) representing the 32-byte binary digest. No prefix, no whitespace, no uppercase.¶
3. AnchorDigest MUST exactly equal MerkleRoot with the sha256: prefix stripped. No rehashing, no encoding transformation — purely prefix removal.¶

TSA Binding Requirements:¶

1. Verifiers MUST extract messageImprint from TSTInfo within the TimeStampToken.¶
2. Verifiers MUST confirm that messageImprint.hashAlgorithm equals the SHA-256 OID (2.16.840.1.101.3.4.2.1). If the OID differs, verification MUST return INVALID.¶
3. Verifiers MUST confirm that the lowercase hexadecimal encoding of messageImprint.hashedMessage (32 bytes) is identical to AnchorDigest. This is a byte-for-byte comparison after hex encoding; case-insensitive comparison is NOT sufficient because AnchorDigest is already required to be lowercase.¶
4. If an implementation transmits AnchorDigest to the TSA as a UTF-8 string rather than a 32-byte binary OCTET STRING, the TSA will timestamp a 64-byte value (the hex encoding) rather than the correct 32-byte digest. This is a known implementation error. Verifiers MUST detect this by comparing the messageImprint length: if hashedMessage is not exactly 32 bytes, verification MUST return INVALID.¶

PROHIBITED Patterns (known implementation errors):¶

• Submitting the hex string of AnchorDigest (64 ASCII bytes) instead of the decoded binary (32 bytes) to the TSA — results in messageImprint mismatch¶
• Computing SHA256(MerkleRoot) as the TSA input — results in double-hashing¶
• Computing SHA256(AnchorDigest_as_string) — results in hashing the hex encoding rather than the binary value¶
• Using the MerkleRoot string (with sha256: prefix) directly as the TSA hashedMessage — results in a 71-byte input instead of 32 bytes¶

function verifyAnchorDigestBinding(anchor):
    // Format checks
    if not matches(anchor.Merkle.Root, /^sha256:[0-9a-f]{64}$/):
        return INVALID("MerkleRoot format violation")

    if not matches(anchor.AnchorDigest, /^[0-9a-f]{64}$/):
        return INVALID("AnchorDigest format violation")

    // Prefix-strip equivalence
    expectedDigest = anchor.Merkle.Root.substring(7)
    if anchor.AnchorDigest != expectedDigest:
        return INVALID("AnchorDigest != MerkleRoot hex part")

    // TSA binding
    tstInfo = extractTSTInfo(anchor.TSA.Token)
    if tstInfo.messageImprint.hashAlgorithm != SHA256_OID:
        return INVALID("TSA hash algorithm is not SHA-256")

    imprintBytes = tstInfo.messageImprint.hashedMessage
    if length(imprintBytes) != 32:
        return INVALID("TSA hashedMessage is not 32 bytes"
                       + " (string-as-binary error?)")

    imprintHex = lowercase_hex(imprintBytes)
    if imprintHex != anchor.AnchorDigest:
        return INVALID("TSA messageImprint != AnchorDigest")

    return VALID

7.6. Chain Integrity Verification

GENESIS_PREV_HASH = "sha256:00000000000000000000000000000000" +
                    "00000000000000000000000000000000"

function verifyChainIntegrity(events):
    if events.length == 0:
        return VALID

    if events[0].PrevHash != GENESIS_PREV_HASH:
        return CHAIN_INTEGRITY_VIOLATION("Invalid genesis PrevHash")

    for i in range(1, events.length):
        expectedPrevHash = events[i-1].EventHash
        if events[i].PrevHash != expectedPrevHash:
            return CHAIN_INTEGRITY_VIOLATION(
                "Break at event " + i)

    return VALID

7.7. Completeness Invariant Verification

function verifyCompleteness(events, sealEvent):
    ci = sealEvent.CompletenessInvariant

    // Step 1: Verify count
    if events.length != ci.ExpectedCount:
        return COMPLETENESS_VIOLATION("Count mismatch")

    // Step 2: Compute XOR hash sum
    computed = bytes(32)
    for event in events:
        eventHashBytes = hexDecode(event.EventHash.substring(7))
        computed = XOR(computed, eventHashBytes)

    // Step 3: Compare with sealed value
    expectedHashSum = hexDecode(ci.HashSum.substring(7))
    if computed != expectedHashSum:
        return COMPLETENESS_VIOLATION("Hash sum mismatch")

    // Step 4: Verify timestamp bounds
    for event in events:
        if event.Timestamp < ci.FirstTimestamp:
            return COMPLETENESS_VIOLATION("Before collection start")
        if event.Timestamp > ci.LastTimestamp:
            return COMPLETENESS_VIOLATION("After collection end")

    return VALID

8.1. DepthAnalysis Structure

When present in CaptureContext, DepthAnalysis MUST include:¶

8.2. Sensor Types

8.3. Depth Analysis Verification

Depth analysis data is included in the EventHash computation, making it tamper-evident. Verifiers MAY use DepthAnalysis to assess capture environment but MUST NOT treat it as definitive proof of scene authenticity. Depth sensors can be spoofed by sophisticated adversaries.¶

When ScreenDetected is true, implementations SHOULD display a warning to users but MUST NOT automatically reject the proof.¶

9.1. Design Principles

• Silent Failure: If verification fails or times out, sharing MUST proceed without any user-visible error or delay.¶
• No Truth Claims: The indicator communicates provenance availability, not content truthfulness.¶
• Performance Budget: Verification MUST complete within 200ms or short-circuit to silent passthrough.¶

9.2. VerificationResult

Only PROVENANCE_AVAILABLE results in visible indication. All other statuses MUST result in silent passthrough where the original content is shared without modification or delay.¶

9.3. Prohibited Terminology

Implementations MUST NOT use the following terms in user-facing displays related to CPP provenance:¶

• "Verified" or "Verified Content"¶
• "Authentic" or "Authenticated"¶
• "True" or "Truthful"¶
• "Certified"¶
• "Guaranteed"¶
• "Real"¶
• "Trustworthy"¶

Recommended terminology: "Provenance Available", "Capture Provenance Recorded", "Origin Traceable".¶

10.1. Field Mapping

10.2. Dual-Standard Output

Implementations generating both CPP and C2PA manifests MUST ensure shared fields (Timestamp, GPS, DeviceInfo) are consistent across both manifests. CPP-specific fields (DepthAnalysis, HumanAttestation, CompletenessInvariant) have no C2PA equivalent and are carried only in the CPP manifest.¶

11.1. Location Data

Location collection SHOULD be disabled by default. When enabled, implementations SHOULD:¶

• Clearly indicate when location is being recorded¶
• Allow users to delete location from individual events¶
• Consider privacy implications of location precision¶
• Support approximate location modes that reduce GPS precision¶

11.2. Biometric Data

Implementations MUST NOT store raw biometric data (fingerprints, face images). Human presence verification, if implemented, SHOULD:¶

• Process biometrics locally on-device¶
• Store only verification results (boolean flags)¶
• Never transmit biometric data to external services¶

11.3. Tombstone Privacy

When events are deleted via TOMBSTONE:¶

• Original event content is removed¶
• TOMBSTONE preserves chain integrity¶
• Reason codes allow selective disclosure¶

11.4. Shareable vs Forensic Proofs

12.1. Hash Algorithm Agility

This specification mandates SHA-256 for all hash computations. Future versions MAY define additional algorithms via the HashAlgo field. Verifiers MUST reject unknown hash algorithms.¶

Post-quantum consideration: ML-DSA-65 (NIST Module-Lattice Digital Signature Algorithm) is RESERVED for future adoption. XOR-based accumulators used in the Completeness Invariant, being purely symmetric operations, face fewer quantum vulnerabilities than public-key constructions.¶

12.2. Signature Algorithm Requirements

Implementations MUST support ES256 (ECDSA with P-256 and SHA-256, per [FIPS186-5]) for mobile device compatibility. Ed25519 MAY be supported for non-mobile implementations.¶

Private keys SHOULD be stored in hardware security modules where available:¶

• iOS: Secure Enclave (kSecAttrTokenIDSecureEnclave)¶
• Android: StrongBox or TEE via Android Keystore¶
• Desktop/Server: TPM 2.0 or HSM¶

Hardware-backed keys provide non-exportability guarantees that strengthen the binding between device and signature.¶

12.3. TSA Trust

Security of timestamp proofs depends on TSA trustworthiness. Implementations:¶

• MUST verify the CMS signature in the TimeStampToken per [RFC5652]¶
• SHOULD validate the certificate chain to a configured trust anchor¶
• SHOULD use TSAs with published certificate policies¶
• MAY support multiple TSA services for redundancy¶

12.4. Merkle Tree Security

The 0x00/0x01 prefix bytes ensure leaf hashes cannot equal internal node hashes for any input. This prevents second preimage attacks on the tree structure. This construction differs from Certificate Transparency [RFC6962] which uses a similar but incompatible scheme.¶

12.5. Clock Accuracy

Device timestamps (Timestamp field) are self-attested and may be inaccurate. The authoritative timestamp is GenTime from the TSA response. Implementations SHOULD warn users when device time differs significantly from TSA GenTime (e.g., more than 5 minutes).¶

12.6. Deletion Detection Limitations

The Completeness Invariant detects deletions within a sealed collection. It does NOT detect:¶

• Events never created (adversary captured but never recorded)¶
• Events in other collections¶
• Deletions before sealing¶

XOR is commutative and self-inverse. An attacker who can forge TWO events with EventHashes that XOR to zero can delete both without detection. The CI is designed to work WITH the Merkle tree anchor, not replace it.¶

12.7. Depth Analysis Security

Depth sensors can be spoofed by sophisticated adversaries using 3D displays or structured light interference. Depth analysis provides an additional signal but is not a definitive screen detection mechanism. Implementations MUST NOT represent depth analysis as conclusive proof of scene authenticity.¶

12.8. Canonicalization Attacks

JSON canonicalization per [RFC8785] prevents ordering and whitespace attacks. Implementations MUST ensure field names exactly match the specification (PascalCase for events) and Unicode normalization is handled consistently.¶

14.1. VeraSnap iOS (Non-Normative)

VeraSnap is a consumer iOS application implementing CPP, available in 175 countries with 10-language localization. It demonstrates:¶

• Secure Enclave key storage with ES256 signatures¶
• RFC 3161 TSA integration with multiple providers (failover)¶
• Merkle tree construction per this specification¶
• Offline proof verification¶
• LiDAR-based depth analysis for screen detection¶
• Cross-platform proof export and QR code sharing¶

14.2. VeraSnap Android (Non-Normative)

A Kotlin-based Android implementation validates cross-platform interoperability:¶

• Android Keystore (StrongBox/TEE) key storage with ES256¶
• RFC 8785 JSON canonicalization producing identical hashes to iOS¶
• Proof JSON verification across platforms¶
• QR code generation and scanning interoperability¶

Cross-platform testing confirmed that proofs generated on iOS verify correctly on Android and vice versa, validating the canonicalization and hashing specifications.¶

A.1. Canonical Event with SignerInfo (Normative)

The canonical event structure uses PascalCase field names. This is the structure that MUST be used for EventHash computation.¶

{
  "EventID": "550e8400-e29b-41d4-a716-446655440001",
  "ChainID": "urn:uuid:550e8400-e29b-41d4-a716-446655440000",
  "PrevHash": "sha256:0000000000000000000000000000000000000000000000000000000000000000",
  "Timestamp": "2026-01-27T10:30:00.000Z",
  "EventType": "INGEST",
  "HashAlgo": "SHA256",
  "SignAlgo": "ES256",
  "Asset": {
    "AssetID": "asset-001",
    "AssetType": "IMAGE",
    "AssetHash": "sha256:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
    "AssetName": "IMG_0001.HEIC",
    "MimeType": "image/heic"
  },
  "SignerInfo": {
    "Name": "John Doe",
    "Identifier": null,
    "AttestedAt": "2026-01-27T10:29:55.000Z"
  },
  "DeviceInfo": {
    "Manufacturer": "Apple",
    "Model": "iPhone 15 Pro",
    "DeviceClass": "SMARTPHONE",
    "OSName": "iOS",
    "OSVersion": "17.3",
    "AppVersion": "1.5.34"
  },
  "EventHash": "sha256:7d865e959b2466918c9863afca942d0fb89d7c9ac0c99bafc3749504ded97730",
  "Signature": "MEUCIQDKsRwMv..."
}

A.2. Anchor Structure with MessageImprint (Normative)

{
  "Anchor": {
    "AnchorID": "anchor-001",
    "AnchorType": "RFC3161",
    "AnchorDigest": "719f871f1018a17ebe199d4f0db27e3a4929f8ab3e46f5c0d30054f4b331e929",
    "AnchorDigestAlgorithm": "sha-256",
    "Merkle": {
      "TreeSize": 1,
      "LeafHashMethod": "SHA256(0x00||EventHash)",
      "LeafHash": "sha256:719f871f1018a17ebe199d4f0db27e3a4929f8ab3e46f5c0d30054f4b331e929",
      "LeafIndex": 0,
      "Proof": [],
      "Root": "sha256:719f871f1018a17ebe199d4f0db27e3a4929f8ab3e46f5c0d30054f4b331e929"
    },
    "TSA": {
      "Token": "MIIEzAYJKoZIhvcNAQcCoIIEvTCCBLkCAQMx...",
      "MessageImprint": {
        "HashAlgorithm": "sha-256",
        "HashedMessage": "719f871f1018a17ebe199d4f0db27e3a4929f8ab3e46f5c0d30054f4b331e929"
      },
      "GenTime": "2026-01-27T10:31:00.000Z",
      "Service": "https://freetsa.org/tsr"
    }
  }
}

A.3. Chain Context Example (Non-Normative)

{
  "ChainContext": {
    "ChainID": "urn:uuid:550e8400-e29b-41d4-a716-446655440000",
    "TotalEvents": 100,
    "ActiveEvents": 53,
    "TombstoneCount": 47,
    "EventPosition": 15,
    "CompletenessInvariant": {
      "ExpectedCount": 100,
      "HashSum": "sha256:a3f2c8d1e5b9...",
      "FirstTimestamp": "2026-01-27T10:30:00.000Z",
      "LastTimestamp": "2026-01-27T17:45:00.000Z"
    },
    "GeneratedAt": "2026-01-27T18:00:00.000Z"
  }
}

A.4. Depth Analysis Example (Non-Normative)

{
  "CaptureContext": {
    "DepthAnalysis": {
      "SensorType": "LIDAR",
      "FlatnessScore": 0.12,
      "DepthVariance": 0.847,
      "ScreenDetected": false,
      "Confidence": 0.95,
      "AnalysisVersion": "1.4.0"
    }
  }
}

B.1. Test Vector 1: Single-Leaf Tree

Input:¶

EventHash = "sha256:7d865e959b2466918c9863afca942d0fb89d7c9ac0c99bafc3749504ded97730"

Computation:¶

LeafHash = SHA256(0x00 || EventHash_bytes)
         = sha256:719f871f1018a17ebe199d4f0db27e3a4929f8ab3e46f5c0d30054f4b331e929

For TreeSize=1:
  Root = LeafHash
  LeafIndex = 0
  Proof = []
  AnchorDigest = 719f871f1018a17ebe199d4f0db27e3a4929f8ab3e46f5c0d30054f4b331e929

Result: VALID

B.2. Test Vector 2: Two-Leaf Tree

EventHash[0] = "sha256:aaaa...aaaa"  (32 bytes of 0xaa)
EventHash[1] = "sha256:bbbb...bbbb"  (32 bytes of 0xbb)

L0 = SHA256(0x00 || 0xaa...aa)
   = sha256:e0bb82791bae3c50bd9c20fa4ccdcb8064a56e5c12bc69b07e6712ac9b4429e6
L1 = SHA256(0x00 || 0xbb...bb)
   = sha256:4f16119d36ccd0da91102f57692d73934fd0ad2494280df88449accedbbfb7ea

Root = SHA256(0x01 || L0 || L1)
     = sha256:03938e2c8f758e6cae443d499b41c899c373eb0c0198bae61796a069f2b05904

For index 0: Proof = [L1]
For index 1: Proof = [L0]

Result: VALID

B.3. Test Vector 3: TSA messageImprint Verification

AnchorDigest = "719f871f1018a17ebe199d4f0db27e3a4929f8ab3e46f5c0d30054f4b331e929"

TSTInfo.messageImprint.hashAlgorithm = SHA-256
TSTInfo.messageImprint.hashedMessage = 0x719f871f...1e929

lowercase_hex(hashedMessage) == AnchorDigest ? YES

Result: VALID