# US 2025/0094546 A1 · Watermarking digital media for content verification with differential detection

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0094546 A1
- Document ID: US20250094546A1
- Status: Published application
- Kind: A1
- Application number: 18/966,712
- Filed: 2024-12-03
- Published: 2025-03-20
- Earliest claimed: 2023-07-21
- Assignees: Trufo Inc.
- Applicants: Trufo Inc.
- Inventors: Yan Huang; Xinyi Han; Jiaxin Guan
- Web page: https://patentgrep.com/patent/US20250094546A1

## CPC

- G06F 21/16: Program or content traceability, e.g. by watermarking
- H04L 63/12: {Applying verification of the received information (cryptographic mechanisms or cryptographic arrangements for data integrity or data verification H04L9/32)}
- H04L 9/0643: {Hash functions, e.g. MD5, SHA, HMAC or f9 MAC}
- H04L 9/3247: {involving digital signatures}
- H04N 21/8358: involving watermark {(protecting executable software by watermarking G06F21/16; image watermarking in general G06T1/0021; watermarks inserted in still images for transmission purposes H04N1/32144; inserting watermarks during video coding H04N19/467)}
- H04L 9/3239: {involving non-keyed hash functions, e.g. modification detection codes [MDCs], MD5, SHA or RIPEMD}
- H04L 9/50: {using hash chains, e.g. blockchains or hash trees}
- H04L 2209/608: Watermarking
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)
- H04N: PICTORIAL COMMUNICATION, e.g. TELEVISION

## Abstract

The present disclosure provides a method of watermarking digital media for subsequent decoding. The method includes receiving a secure content package comprising digital audiovisual content, source entity information, content annotations, and digital signatures. The package is validated using public-key cryptography. A content identifier is generated and embedded as a watermark into the digital audiovisual content. The watermark is configured to be machine readable and visually inconspicuous. A content database entry is recorded in a remote database, retrievable using the content identifier and comprising the content annotations and source information. The watermarked digital audiovisual content is stored and transmitted to the media source. The method may further include generating and storing content hashes for future authentication comparisons.

## Claims

### Claim 1

A computer-implemented method of authenticating digital content, comprising:
(a) receiving, by one or more processors, data representing an observed item of digital content;
(b) receiving, by the one or more processors, information comprising a differential hash associated with an original version of said observed item of digital content;
(c) computing, by the one or more processors, a differential hash of said observed item of digital content, wherein the method of computation comprises:
(i) determining a plurality of spatial regions of said observed item of digital content,
(ii) applying a numerical function to each spatial region of the plurality of spatial regions, wherein the numerical function is a computational operation that takes content data from a spatial region as input and produces a numerical output that characterizes or represents that region in a way suitable for subsequent comparison and analysis;
(iii) quantizing respective outputs of the numerical functions to generate quantized outputs, and
(iv) concatenating the quantized outputs into a final hash;
(d) comparing, by the one or more processors, said received differential hash and said computed differential hash by identifying locations at which they differ;
(e) identifying, by the one or more processors, spatial regions of the observed item of digital content that exhibit differences exceeding a predetermined threshold or spatial regions that exhibit differences below said predetermined threshold versus the original version by analyzing a result of said comparing;
(f) generating, by the one or more processors, information comprising said identification of spatial regions; and
(g) storing, in a non-transitory computer-readable medium, the information comprising said identification of spatial regions.

### Claim 2 (depends on 1)

The method of claim 1, wherein:
(a) the received information further comprises a digital signature that identifies a producer of the information as a trustworthy entity registered with a public key infrastructure, and
(b) the method further comprises validating legitimacy of the received differential hash by verifying said digital signature using a public key associated with the trustworthy entity.

### Claim 3 (depends on 1)

The method of claim 1, wherein:
(a) the computation of the differential hash is designed so that an accurate reconstruction of the digital content on which the computation is performed cannot be easily recovered from the differential hash, thereby ensuring a degree of privacy; and
(b) the computation of the differential hash is designed so that a length of the final hash is significantly shorter than a result of a typical content compression algorithm applied to the observed item of digital content.

### Claim 4 (depends on 1)

The method of claim 1, further comprising:
(a) prior to the computing of the differential hash, aligning the observed item of digital content to correct for geometric modifications, so that the determined spatial regions of said observed item of digital content are consistent with corresponding spatial regions of the original version, wherein:
(b) the observed item of digital content contains an embedded watermark;
(c) the received information further comprises one or more alignment hashes; and
(d) the aligning comprises:
(i) extracting geometric information from said embedded watermark, and
(ii) comparing the one or more received alignment hashes with one or more corresponding computed alignment hashes.

### Claim 5 (depends on 1)

The method of claim 1, wherein:
(a) the plurality of spatial regions are determined so that a geometric diameter of each spatial region of the plurality of spatial regions is no more than a specified small fraction of dimensions of the digital content;
(b) the plurality of spatial regions are determined so that each spatial region overlaps with a number of other spatial regions, such that an adjacency graph defined by overlaps has a low diameter and a consistent degree per vertex; and
(c) the numerical function applied to each spatial region is a pseudorandom function generated with a seed that differs between regions.

### Claim 6 (depends on 1)

The method of claim 1, wherein:
(a) the numerical function applied to each spatial region is a perceptive hash designed so that a corresponding quantized output is likely to change if the spatial region is materially changed and not likely to change if the spatial region is materially unchanged; and
(b) said numerical function belongs to a plurality of function classes that each distinguish a different set of material changes.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
(a) prior to the quantizing and the concatenating, transforming a first vector consisting of the respective outputs of the numerical functions into a second vector of numerical outputs through a pre-defined sequence of matrix operations.

### Claim 8 (depends on 1)

The method of claim 1, wherein:
(a) the received information further comprises a differential hash parameter; and
(b) the computing of the differential hash utilizes said differential hash parameter to ensure that the method of computation deterministically matches a method of computation with which the received differential hash was computed on the original version of said observed item of digital content.

### Claim 9 (depends on 1)

The method of claim 1, wherein:
(a) the identifying of spatial regions comprises generating one or more numerical values for each spatial region, wherein each numerical value quantifies an aspect of material change in the respective spatial region.

### Claim 10 (depends on 1)

The method of claim 1, wherein:
(a) the identifying of spatial regions comprises generating a human-interpretable classification of a type of material change in each spatial region;
(b) the analyzing of the result of said comparing further comprises inputting the observed item of digital content alongside the result of said comparing; and
(c) the analyzing utilizes a pre-trained neural network with at least one convolution or attention layer.

### Claim 11

A system for authenticating digital content, comprising:
(a) one or more processors; and
(b) a non-transitory computer-readable medium storing instructions that, when executed by the one or more processors, cause the system to perform operations comprising:
(i) receiving data representing an observed item of digital content;
(ii) receiving information comprising a differential hash associated with an original version of said observed item of digital content;
(iii) computing a differential hash of said observed item of digital content, wherein the method of computation comprises:
(i) determining a plurality of spatial regions of said observed item of digital content,
(ii) applying a numerical function to each spatial region of the plurality of spatial regions, wherein the numerical function is a computational operation that takes content data from a spatial region as input and produces a numerical output that characterizes or represents that region in a way suitable for subsequent comparison and analysis;
(iii) quantizing respective outputs of the numerical functions to generate quantized outputs, and
(iv) concatenating the quantized outputs into a final hash;
(iv) comparing said received differential hash and said computed differential hash by identifying locations at which they differ;
(v) identifying spatial regions of the observed item of digital content that exhibit differences exceeding a predetermined threshold or spatial regions that exhibit differences below said predetermined threshold versus the original version by analyzing a result of said comparing;
(vi) generating information comprising said identification of spatial regions; and
(vii) storing the information comprising said identification of spatial regions.

### Claim 12 (depends on 11)

The system of claim 11, wherein:
(a) the received information further comprises a digital signature that identifies a producer of the information as a trustworthy entity registered with a public key infrastructure, and
(b) the operations further comprise validating legitimacy of the received differential hash by verifying said digital signature using a public key associated with the trustworthy entity.

### Claim 13 (depends on 11)

The system of claim 11, wherein:
(a) the computation of the differential hash is designed so that an accurate reconstruction of the digital content on which the computation is performed cannot be easily recovered from the differential hash, thereby ensuring a degree of privacy; and
(b) the computation of the differential hash is designed so that a length of the final hash is significantly shorter than a result of a typical content compression algorithm applied to the observed item of digital content.

### Claim 14 (depends on 11)

The system of claim 11, wherein the operations further comprise:
(a) prior to the computing of the differential hash, aligning the observed item of digital content to correct for geometric modifications, so that the determined spatial regions of said observed item of digital content are consistent with corresponding spatial regions of the original version, wherein:
(b) the observed item of digital content contains an embedded watermark;
(c) the received information further comprises one or more alignment hashes; and
(d) the aligning comprises:
(i) extracting geometric information from said embedded watermark, and
(ii) comparing the one or more received alignment hashes with one or more corresponding computed alignment hashes.

### Claim 15 (depends on 11)

The system of claim 11, wherein:
(a) the plurality of spatial regions are determined so that a geometric diameter of each spatial region of the plurality of spatial regions is no more than a specified small fraction of dimensions of the digital content;
(b) the plurality of spatial regions are determined so that each spatial region overlaps with a number of other spatial regions, such that an adjacency graph defined by overlaps has a low diameter and a consistent degree per vertex; and
(c) the numerical function applied to each spatial region is a pseudorandom function generated with a seed that differs between regions.

### Claim 16 (depends on 11)

The system of claim 11, wherein:
(a) the numerical function applied to each spatial region is a perceptive hash designed so that a corresponding quantized output is likely to change if the spatial region is materially changed and not likely to change if the spatial region is materially unchanged; and
(b) said numerical function belongs to a plurality of function classes that each distinguish a different set of material changes.

### Claim 17 (depends on 11)

The system of claim 11, wherein the operations further comprise:
(a) prior to the quantizing and the concatenating, transforming a first vector consisting of the respective outputs of the numerical functions into a second vector of numerical outputs through a pre-defined sequence of matrix operations.

### Claim 18 (depends on 11)

The system of claim 11, wherein:
(a) the received information further comprises a differential hash parameter; and
(b) the computing of the differential hash utilizes said differential hash parameter to ensure that the method of computation deterministically matches a method of computation with which the received differential hash was computed on the original version of said observed item of digital content.

### Claim 19 (depends on 11)

The system of claim 11, wherein:
(a) the identifying of spatial regions comprises generating one or more numerical values for each spatial region, wherein each numerical value quantifies an aspect of material change in the respective spatial region.

### Claim 20 (depends on 11)

The system of claim 11, wherein:
(a) the identifying of spatial regions comprises generating a human-interpretable classification of a type of material change in each spatial region;
(b) the analyzing of the result of said comparing further comprises inputting the observed item of digital content alongside the result of said comparing; and
(c) the analyzing utilizes a pre-trained neural network with at least one convolution or attention layer.

### Claim 21

A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations for authenticating digital content, the operations comprising:
(a) receiving data representing an observed item of digital content;
(b) receiving information comprising a differential hash associated with an original version of said observed item of digital content;
(c) computing a differential hash of said observed item of digital content, wherein the method of computation comprises:
(i) determining a plurality of spatial regions of said observed item of digital content,
(ii) applying a numerical function to each spatial region of the plurality of spatial regions, wherein the numerical function is a computational operation that takes content data from a spatial region as input and produces a numerical output that characterizes or represents that region in a way suitable for subsequent comparison and analysis;
(iii) quantizing respective outputs of the numerical functions to generate quantized outputs, and
(iv) concatenating the quantized outputs into a final hash;
(d) comparing said received differential hash and said computed differential hash by identifying locations at which they differ;
(e) identifying spatial regions of the observed item of digital content that exhibit differences exceeding a predetermined threshold or spatial regions that exhibit differences below said predetermined threshold versus the original version by analyzing a result of said comparing;
(f) generating information comprising said identification of spatial regions; and
(g) storing the information comprising said identification of spatial regions.

### Claim 22 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein:
(a) the received information further comprises a digital signature that identifies a producer of the information as a trustworthy entity registered with a public key infrastructure, and
(b) the operations further comprise validating legitimacy of the received differential hash by verifying said digital signature using a public key associated with the trustworthy entity.

### Claim 23 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein:
(a) the computation of the differential hash is designed so that an accurate reconstruction of the digital content on which the computation is performed cannot be easily recovered from the differential hash, thereby ensuring a degree of privacy; and
(b) the computation of the differential hash is designed so that a length of the final hash is significantly shorter than a result of a typical content compression algorithm applied to the observed item of digital content.

### Claim 24 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein the operations further comprise:
(a) prior to the computing of the differential hash, aligning the observed item of digital content to correct for geometric modifications, so that the determined spatial regions of said observed item of digital content are consistent with corresponding spatial regions of the original version, wherein:
(b) the observed item of digital content contains an embedded watermark;
(c) the received information further comprises one or more alignment hashes; and
(d) the aligning comprises:
(i) extracting geometric information from said embedded watermark, and
(ii) comparing the one or more received alignment hashes with one or more corresponding computed alignment hashes.

### Claim 25 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein:
(a) the plurality of spatial regions are determined so that a geometric diameter of each spatial region of the plurality of spatial regions is no more than a specified small fraction of dimensions of the digital content;
(b) the plurality of spatial regions are determined so that each spatial region overlaps with a number of other spatial regions, such that an adjacency graph defined by overlaps has a low diameter and a consistent degree per vertex; and
(c) the numerical function applied to each spatial region is a pseudorandom function generated with a seed that differs between regions.

### Claim 26 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein:
(a) the numerical function applied to each spatial region is a perceptive hash designed so that a corresponding quantized output is likely to change if the spatial region is materially changed and not likely to change if the spatial region is materially unchanged; and
(b) said numerical function belongs to a plurality of function classes that each distinguish a different set of material changes.

### Claim 27 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein the operations further comprise:
(a) prior to the quantizing and the concatenating, transforming a first vector consisting of the respective outputs of the numerical functions into a second vector of numerical outputs through a pre-defined sequence of matrix operations.

### Claim 28 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein:
(a) the received information further comprises a differential hash parameter; and
(b) the computing of the differential hash utilizes said differential hash parameter to ensure that the method of computation deterministically matches a method of computation with which the received differential hash was computed on the original version of said observed item of digital content.

### Claim 29 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein:
(a) the identifying of spatial regions comprises generating one or more numerical values for each spatial region, wherein each numerical value quantifies an aspect of material change in the respective spatial region.

### Claim 30 (depends on 21)

The non-transitory computer-readable medium of claim 21, wherein:
(a) the identifying of spatial regions comprises generating a human-interpretable classification of a type of material change in each spatial region;
(b) the analyzing of the result of said comparing further comprises inputting the observed item of digital content alongside the result of said comparing; and
(c) the analyzing utilizes a pre-trained neural network with at least one convolution or attention layer.
