C2PA v2.1 Updates - BMFF hash v3 assertion (#213)

* Implement BMFF hash v3 assertion

* Enhance BMFF functionality with new box retrieval method and hash assertion improvements

* Refactor BMFFHashAssertion to support versioning and improve validation logic
Add unit tests

* remove unused code

* Add unit tests for v2 and v3 BMFF hash assertions

* add new v3 test cases, add a new asset type for 'trustnxt-icon-signed-v3.heic' in asset-reading tests.

* Refactor AssertionUtils and BMFFHashAssertion for improved exclusion handling and validation logic

- Simplified exclusion sorting in AssertionUtils to only consider start positions.
- Enhanced data processing in AssertionUtils to correctly handle offset markers and remaining data.
- Updated BMFFHashAssertion to use mdatBox.payloadOffset for data retrieval, ensuring compliance with specifications.
- Modified test cases to reflect changes in exclusion structure and improve validation checks for BMFF hash assertions.

* Implement HashAssertion interface in BMFFHashAssertion and DataHashAssertion classes

* bmff hash update logic

* Refactor BMFF hash assertion and box handling

* fix version

* resolve comments

* update readme

* update readme

.changeset/tiny-buses-tease.md

@@ -0,0 +1,5 @@
+---
+'@trustnxt/c2pa-ts': minor
+---
+
+BMFF hash v3 assertion

README.md

@@ -4,7 +4,7 @@
 
 ## About
 
-`c2pa-ts` is a pure TypeScript implementation of [Coalition for Content Provenance and Authenticity (C2PA)](https://c2pa.org/) according to [specification version 2.0](https://c2pa.org/specifications/specifications/2.0/specs/C2PA_Specification.html).
+`c2pa-ts` is a pure TypeScript implementation of [Coalition for Content Provenance and Authenticity (C2PA)](https://c2pa.org/) according to [specification version 2.1](https://c2pa.org/specifications/specifications/2.1/specs/C2PA_Specification.html).
 
 It does not use any native binaries or WebAssembly and is therefore truly platform independent. In modern browsers as well as Node.js it should run out of the box. In mobile apps or other environments lacking browser APIs, some external code may be necessary (see [below](#usage-in-constrained-environments) for details).
 
@@ -24,7 +24,7 @@ Anything that's not listed below is not currently planned to be implemented.
 - :construction: Validating manifests (mostly implemented except chain of trust validation)
 - :white_check_mark: Creating manifests
 
-:information_source: On C2PA versions: The library is targeted at C2PA specification 2.0, however data structures from older versions of the specification are also supported for backwards compatibility.
+:information_source: On C2PA versions: The library is targeted at C2PA specification 2.1, however data structures from older versions of the specification are also supported for backwards compatibility.
 
 :information_source: Although it is a separate project from C2PA, the library also includes support for several [CAWG](https://github.com/creator-assertions/) assertions.
 
@@ -36,15 +36,16 @@ Anything that's not listed below is not currently planned to be implemented.
 - :x: GIF
 - :x: TIFF
 - :x: WebP
+- :x: JPEG XL
 
 ### Supported assertions
 
 - :white_check_mark: Data Hash
-- :white_check_mark: BMFF-Based Hash (except Merkle tree hashing)
+- :white_check_mark: BMFF-Based Hash (v2 and v3)
 - :x: General Boxes Hash
 - :white_check_mark: Thumbnail
 - :white_check_mark: Actions (except action templates and metadata)
-- :white_check_mark: Ingredient
+- :white_check_mark: Ingredient (v2 and v3)
 - :white_check_mark: Metadata (specialized, common, generic, and CAWG variants)
 - :white_check_mark: Creative Work
 - :white_check_mark: Training and Data Mining (C2PA and CAWG variants)

src/asset/BMFF.ts

@@ -3,6 +3,10 @@ import { BaseAsset } from './BaseAsset';
 import { Asset } from './types';
 
 export class BMFF extends BaseAsset implements Asset {
+    public static c2paBoxUserType = [
+        0xd8, 0xfe, 0xc3, 0xd6, 0x1b, 0x0e, 0x48, 0x3c, 0x92, 0x97, 0x58, 0x28, 0x87, 0x7e, 0xc4, 0x81,
+    ];
+
     /** Currently supported major brand identifiers */
     private static canReadBrands = new Set(['heic', 'mif1']);
 
@@ -234,7 +238,7 @@ class BoxReader {
 
             // Now that the box header is fully read, we know that it might be a UUID box (== has a userType),
             // so handle those cases as well (currently only C2PABox)
-            if (box.userType && BinaryHelper.bufEqual(box.userType, C2PABox.c2paUserType)) {
+            if (box.userType && BinaryHelper.bufEqual(box.userType, BMFF.c2paBoxUserType)) {
                 box = new C2PABox(pos, size, payloadPos, payloadSize, boxType);
                 box.readContents(buf);
             }
@@ -578,10 +582,6 @@ interface C2PAManifestBoxPayload extends C2PABoxPayload {
 }
 
 class C2PABox extends FullBox<C2PABoxPayload> {
-    public static c2paUserType = [
-        0xd8, 0xfe, 0xc3, 0xd6, 0x1b, 0x0e, 0x48, 0x3c, 0x92, 0x97, 0x58, 0x28, 0x87, 0x7e, 0xc4, 0x81,
-    ];
-
     private static readonly headerLength =
         4 + // size
         4 + // type
@@ -630,7 +630,7 @@ class C2PABox extends FullBox<C2PABoxPayload> {
             'uuid',
         );
 
-        box.userType = new Uint8Array(C2PABox.c2paUserType);
+        box.userType = new Uint8Array(BMFF.c2paBoxUserType);
         const payload: C2PAManifestBoxPayload = {
             version: 0,
             flags: 0,

src/manifest/AssertionStore.ts

@@ -69,7 +69,9 @@ export class AssertionStore implements ManifestComponent {
         if (label.label === AssertionLabels.actions || label.label === AssertionLabels.actionsV2) {
             assertion = new ActionAssertion();
         } else if (label.label === AssertionLabels.bmffV2Hash) {
-            assertion = new BMFFHashAssertion();
+            assertion = new BMFFHashAssertion(2);
+        } else if (label.label === AssertionLabels.bmffV3Hash) {
+            assertion = new BMFFHashAssertion(3);
         } else if (label.label === AssertionLabels.creativeWork) {
             assertion = new CreativeWorkAssertion();
         } else if (label.label === AssertionLabels.dataHash) {

src/manifest/assertions/AssertionLabels.ts

@@ -4,13 +4,15 @@ export class AssertionLabels {
     public static readonly collectionHash = 'c2pa.hash.collection.data';
     public static readonly bmffHash = 'c2pa.hash.bmff';
     public static readonly bmffV2Hash = 'c2pa.hash.bmff.v2';
+    public static readonly bmffV3Hash = 'c2pa.hash.bmff.v3';
     public static readonly hardBindings = [
         AssertionLabels.dataHash,
         AssertionLabels.boxHash,
         AssertionLabels.collectionHash,
         // "A validator or consumer shall not validate content authenticated by a c2pa.hash.bmff assertion. Instead, it shall report the content as unauthenticated, as if no manifest were present."
         //AssertionLabels.bmffHash,
         AssertionLabels.bmffV2Hash,
+        AssertionLabels.bmffV3Hash,
     ];
 
     public static readonly ingredient = 'c2pa.ingredient';

src/manifest/assertions/AssertionUtils.ts

@@ -14,36 +14,33 @@ export class AssertionUtils {
             return Crypto.digest(await asset.getDataRange(), algorithm);
         }
 
-        // Sort exclusions by start, however make sure offset markers appear first
-        exclusions.sort((a, b) => {
-            const startDiff = a.start - b.start;
-            if (startDiff !== 0) return startDiff;
-            if (a.offsetMarker && !b.offsetMarker) return -1;
-            if (!a.offsetMarker && b.offsetMarker) return 1;
-            return 0;
-        });
+        // Sort exclusions by start position only
+        exclusions.sort((a, b) => a.start - b.start);
 
         const digest = Crypto.streamingDigest(algorithm);
+        let currentPosition = 0;
 
-        for (let i = 0; i < exclusions.length; i++) {
-            const previousEnd = i > 0 ? exclusions[i - 1].start + exclusions[i - 1].length : 0;
-            const length = exclusions[i].start - previousEnd;
+        for (const exclusion of exclusions) {
+            // Write data up to this position
+            if (exclusion.start > currentPosition) {
+                digest.update(await asset.getDataRange(currentPosition, exclusion.start - currentPosition));
+            }
 
-            if (exclusions[i].offsetMarker) {
+            // Handle offset markers
+            if (exclusion.offsetMarker) {
                 const offsetBytes = new Uint8Array(8);
                 const view = new DataView(offsetBytes.buffer);
-                view.setBigInt64(0, BigInt(exclusions[i].start), false);
+                view.setBigInt64(0, BigInt(exclusion.start), false);
                 digest.update(offsetBytes);
-            }
-
-            if (length > 0) {
-                digest.update(await asset.getDataRange(previousEnd, length));
+                currentPosition = exclusion.start; // Don't skip any data for offset markers
+            } else {
+                currentPosition = exclusion.start + exclusion.length;
             }
         }
 
-        const endOfLastExclusion = exclusions[exclusions.length - 1].start + exclusions[exclusions.length - 1].length;
-        if (asset.getDataLength() > endOfLastExclusion) {
-            digest.update(await asset.getDataRange(endOfLastExclusion));
+        // Hash any remaining data
+        if (currentPosition < asset.getDataLength()) {
+            digest.update(await asset.getDataRange(currentPosition));
         }
 
         return digest.final();