Fingerprint Transforming for Unauthorized Media Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Video content providers face difficulties in detecting unauthorized media content items, particularly three-dimensional video content that has been modified to evade detection by altering playback speed or other transformations, making it challenging to match with two-dimensional reference files in databases.

Innovation Solution

A method and system that generate and compare fingerprints of transformed media content items based on abuse queries and transforms, using playback information and fingerprinting techniques to identify matches with stored reference content, allowing continuous detection of circumvention techniques and unauthorized content.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional fingerprinting techniques are used to detect unauthorized content, then detection simplicity is maintained, but detection accuracy deteriorates when content is transformed (e.g., 3D video from 2D, playback speed changes)

Engineering Contradiction:
Improvedetection accuracyVSAvoiddetection system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system pre-generates multiple transformed versions of reference content (e.g., converting 2D reference video to 3D, creating different playback speed versions) and pre-computes their fingerprints before detection occurs. When detecting unauthorized content, the system compares uploaded content against these pre-prepared transformed references, enabling accurate detection of transformed content without real-time transformation complexity.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system varies fingerprinting parameters and transformation parameters to match different circumvention techniques. By adjusting transformation parameters (e.g., 3D conversion methods, playback speed adjustments) and generating corresponding fingerprints, the system adapts to detect various transformed content types while maintaining a unified detection framework.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If multiple transformed versions of reference content are generated to detect circumvention techniques, then detection coverage is improved, but processing time increases

Engineering Contradiction:
Improvedetection coverageVSAvoidprocessing time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

Multiple transformed versions of reference content and their fingerprints are generated in advance during an offline preparation phase. This pre-computation allows the detection system to quickly compare uploaded content against the pre-generated library of transformed references during online detection, achieving comprehensive detection coverage without real-time processing delays.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system periodically updates and regenerates transformed reference content and their fingerprints at scheduled intervals rather than continuously. This periodic regeneration maintains detection coverage for new circumvention techniques while avoiding constant processing overhead, balancing versatility with time efficiency.

Inventive Principle:
Principle #19Periodic action

3Reliability

If fingerprint transformation is applied to match transformed content, then detection capability against circumvention techniques is improved, but computational complexity increases

Engineering Contradiction:
Improvedetection capabilityVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

Instead of transforming the uploaded content for comparison, the system creates fingerprint copies of transformed reference content and compares these pre-computed fingerprints against the uploaded content's fingerprint. This copying approach maintains detection capability for transformed content while avoiding the computational burden of real-time transformation and re-fingerprinting.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The system inverts the traditional detection approach by transforming the reference content to match potential circumvention techniques rather than transforming the uploaded content for comparison. This inversion shifts the computational complexity to an offline preparation phase, enabling reliable detection with simpler online processing.

Inventive Principle:
Principle #13The other way round (Inversion)

Data Source

PatentEP3596628B1Methods, systems and media for transforming fingerprints to detect unauthorized media content items
Publication Date: 2023.11.08 GOOGLE LLC
  • EP3596628B1 patent drawingFigure 1~2
  • EP3596628B1 patent drawingFigure 3
  • EP3596628B1 patent drawingFigure 4

AI summary

Methods, systems, and media for transforming fingerprints to detect unauthorized media content items are provided. The method comprises: receiving criteria relating to an application of a circumvention technique to one or more video content items, wherein the criteria includes abuse criteria that describes the circumvention technique and a transform for use with the one or more video content items in which the circumvention technique was applied; generating an abuse query that includes at least a portion of the abuse criteria that describes the circumvention technique; determining from a plurality of video content items, a subset of video content items responsive to the abuse query; applying, for each video content item in the subset of video content items, the transform to each video content item to obtain a transformed video content item; generating, for each transformed video content item, a fingerprint that represents the transformed video content item; and comparing the fingerprint of the transformed video content item to a plurality of fingerprints associated with reference video content items to determine whether the video content item corresponding to the transformed video content item matches one of the reference video content items.