Object-Specific Image Text Recognition for Data Loss Prevention

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Organizations face challenges in tracking and preventing the leakage of sensitive information, particularly in images, due to the limitations of conventional optical character recognition (OCR) technologies, which are inefficient and unable to accurately handle images with poor quality, rotation, or multiple objects, leading to high computing resource costs and latency.

Innovation Solution

A computer-implemented method that uses computer vision techniques, object-specific detection, and text-based matching to identify objects in images, extract text from specific regions, and evaluate whether the content contains sensitive information, thereby extending data loss prevention (DLP) systems to images with low latency and high efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional OCR technologies are used to analyze images for sensitive information, then text recognition capability is provided, but computing resource cost and latency increase substantially

Engineering Contradiction:
Improvetext recognition accuracyVSAvoidprocessing speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent segments the image analysis process into distinct stages: object detection first, then text recognition only in specific regions of interest. This segmentation avoids applying computationally intensive OCR to entire images, thereby reducing processing time and resource consumption while maintaining text recognition accuracy in relevant areas.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and focuses analysis only on specific regions containing objects of interest rather than processing the entire image. By taking out only the relevant portions for text recognition, the system significantly reduces computing resource requirements and latency while preserving the ability to accurately recognize text where it matters.

Inventive Principle:
Principle #2Taking out (Extraction)

2Measurement precision

If conventional OCR technologies are used to analyze images, then text content can be identified, but the system cannot handle images with poor quality, rotation, or multiple objects efficiently

Engineering Contradiction:
Improvecontent detection accuracyVSAvoidimage condition handling capability
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent performs preliminary object detection and classification before text recognition. This preliminary action identifies objects of interest, determines their orientation and position, and prepares appropriate processing parameters. This preparatory step enables the system to handle rotated images, poor quality images, and multiple objects efficiently by adapting the text recognition process to each detected object's characteristics.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements dynamic adaptation to various image conditions by adjusting processing parameters based on detected object characteristics. The system dynamically handles rotation by detecting object orientation, manages poor quality through selective region processing, and handles multiple objects by iteratively processing each detected object. This dynamic approach maintains detection accuracy across diverse and challenging image conditions.

Inventive Principle:
Principle #15Dynamics

3Reliability

If comprehensive image analysis is performed to ensure accurate detection of sensitive information, then detection reliability improves, but processing time and computing resources increase

Engineering Contradiction:
Improvesensitive information detection reliabilityVSAvoidprocessing latency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent extracts and processes only specific regions containing detected objects rather than performing comprehensive analysis of the entire image. This extraction approach maintains detection reliability for sensitive information by focusing computational resources on relevant areas, while significantly reducing processing time and resource consumption by excluding irrelevant regions.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by performing text recognition only in specific regions where objects are detected, rather than analyzing the entire image. This partial processing approach maintains sufficient detection reliability for sensitive information while reducing overall processing time and computing resource requirements compared to exhaustive full-image analysis.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3574449B1Structured text and pattern matching for data loss prevention in object-specific image domain
Publication Date: 2023.09.06 CA TECH INC
  • EP3574449B1 patent drawingFigure 1
  • EP3574449B1 patent drawingFigure 2
  • EP3574449B1 patent drawingFigure 3

AI summary

Structured text and pattern matching may be performed for data loss prevention in object-specific image domain. According to some embodiments, a method may include receiving an image, identifying one or more objects in the image based on attributes of the one or more objects, and determining an object type of a first object of the one or more objects by a computing device. The method may include identifying, by the computing device, one or more specific regions of the first object for recognition based on the object type of the first object and recognizing text in the one or more specific regions of the first object. In some embodiments, the method may then include providing, by the computing device, the text recognized in the one or more specific regions of the first object to a security engine, wherein the security engine may be configured to evaluate whether the text comprises sensitive information.