Localized Anomaly Detection in Digital Documents via Machine Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional anomaly detection methods in digital documents are inefficient and require structured data, making it difficult to detect and localize modifications in unstructured digital documents.

Innovation Solution

A computer-implemented method using a machine learning framework that involves identifying images in digital documents, processing them with an image compression algorithm, and applying two machine learning models: one to detect modifications and another to localize them within the images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If conventional anomaly detection methods are used, then detection can be performed, but the methods are inefficient and require structured data which limits applicability to unstructured digital documents

Engineering Contradiction:

Improveapplicability to unstructured documentsVSAvoiddetection efficiency

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent transforms the input data representation from structured formats to image-based unstructured data, enabling the detection system to process digital documents regardless of their original structure. This parameter change in data representation allows the system to handle diverse document types while maintaining detection efficiency through specialized image processing algorithms.

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If image processing is applied to detect modifications, then localization accuracy is improved, but processing time and computational resources increase

Engineering Contradiction:

Improvemodification localization accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent divides the image processing task into distinct segments: a classification model that first determines whether modifications exist, and a localization model that then identifies specific modified regions. This segmentation allows the system to avoid full localization processing for all images, reducing overall processing time while maintaining high localization accuracy when modifications are detected.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The classification model performs preliminary action by pre-screening images to identify those containing modifications before applying the more computationally intensive localization model. This preliminary classification step filters out unchanged images, significantly reducing the number of images requiring full localization analysis and thereby decreasing total processing time.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12277792B2Localized anomaly detection in digital documents using machine learning techniques

Publication Date: 2025.04.15 DELL PROD LP

AI summary

Methods, apparatus, and processor-readable storage media for localized anomaly detection in digital documents are provided herein. An example computer-implemented method includes identifying at least one image in a digital document; processing the identified at least one image using at least an image compression algorithm; applying a first machine learning model to the at least one processed image, wherein the first machine learning model is trained to detect whether the at least one image comprises one or more modifications; in response to detecting that the at least one image comprises at least one modification, applying a second machine learning model to identify a location in the at least one image corresponding to the at least one modification; and generating an indication that identifies the location of the at least one modification in the at least one image.