Unwanted Data Detection via Rendered Format Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies are ineffective in detecting unwanted data, such as spam and pornographic material, due to techniques used by generators to circumvent detection, including hidden text and images, which automated systems struggle to identify.
Innovation Solution
A system utilizing a neural network or optical character recognition (OCR) to render and analyze data, specifically detecting indicators of unwanted content by converting graphical data into character codes and using trained neural networks to identify patterns associated with unwanted data, thereby distinguishing between visible and hidden text.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional algorithms use word groupings and header information with Bayesian logic to detect unwanted data, then detection can be performed using automated mechanisms, but generators can circumvent detection by using hidden text, images, or distorted formatting that automated systems struggle to identify
Solution Approach 1:
The patent replaces traditional text-based analysis algorithms with optical recognition technology. Instead of parsing text headers and word groupings, the system captures images of the data display and uses optical character recognition to detect unwanted content. This substitution allows the system to see exactly what a human user sees, including hidden text and images that traditional algorithms miss.
Solution Approach 2:
The patent introduces an intermediary step between data reception and analysis: rendering the data as a visual image. By converting the displayed data into an image format and then analyzing that image through OCR, the system creates a mediator that bridges the gap between raw data and meaningful detection, allowing visual inspection of content that would otherwise be hidden or difficult to analyze.
2Ease of operation
If generators use small words in small font or pictures to convey unwanted messages, then the unwanted data becomes difficult to detect by automated mechanisms, but human readers can still read the message
Solution Approach 1:
The patent replaces automated text parsing with optical image recognition. By capturing the visual display and using OCR to recognize characters in the image, the system can detect small text and images exactly as they appear to human readers, eliminating the advantage that generators have against traditional automated detection systems.
Solution Approach 2:
Instead of trying to make the automated system understand human-readable formats like small text and images, the patent inverts the approach: it renders the display as an image and then analyzes that image. This inversion allows the system to work with visual information directly, making it equally effective at detecting both human-readable and machine-generated content.
3Adaptability or versatility
If legitimate-appearing text is included in near white color on white background with DHTML overlay, then the unwanted message can distract or circumvent detection technology, but the legitimate text provides a cover
Solution Approach 1:
The patent replaces algorithmic text analysis with optical image capture and recognition. By taking a visual snapshot of the display and analyzing the image, the system can see through DHTML overlays and color-matched text because it is analyzing the actual visual rendering rather than parsing source code or text headers. This allows detection of the true content behind deceptive formatting.
Solution Approach 2:
The patent creates a visual copy (image) of the displayed data and analyzes that copy instead of the original data structure. This copying approach allows the system to examine the final rendered output, including any overlays or color-matched text, in its actual visual form, making it possible to detect unwanted content that is hidden through formatting tricks.
Data Source
AI summary
A system, method and computer program product are provided for detecting unwanted data. In use, data is rendered, after which it may be determined whether the rendered data is unwanted, utilizing either a neural network or optical character recognition.