Digital Document Tampering Detection via Multi-Model Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anti-tampering techniques for digital documents have blind spots that limit their ability to detect sophisticated tampering methods, such as image retouching or image blending, and are often unreliable due to dependencies on specific blurring filters.
Innovation Solution
Combining active and passive tampering detection techniques using multiple models to deploy various anti-tampering methods in conjunction, including passive tamper detection engines that analyze grayscale versions of documents and active tamper detection engines that compare RGB values to historical sets, along with unique digital marker analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a singular anti-tampering technique (e.g., LBP-based or double blurring correlation) is used, then the detection process is simple, but the ability to detect sophisticated tampering (e.g., image retouching or blending) is limited
Solution Approach 1:
The patent combines multiple anti-tampering techniques (LBP-based detection, double blurring correlation, and singular value decomposition) into a unified detection system. Each technique processes the document independently and their results are aggregated to produce a final tampering assessment, thereby improving detection reliability while managing system complexity through modular architecture
Solution Approach 2:
The detection system uses a composite approach by integrating multiple detection algorithms with different strengths. LBP handles copy-paste detection, double blurring handles insertion/deletion, and SVD handles sophisticated retouching. This composite methodology ensures comprehensive coverage of various tampering types
2Reliability
If double blurring correlation methods are used, then certain tampering can be detected, but the reliability depends on specific blurring filter types (e.g., gaussian, median)
Solution Approach 1:
The system implements multiple blurring filter types (Gaussian, median, and other filters) within the double blurring correlation methodology. Each filter type is applied in different detection scenarios, allowing the system to adapt to various document characteristics and tampering methods without being constrained to a single filter type
Solution Approach 2:
The patent adjusts blurring filter parameters (kernel size, sigma values for Gaussian, median window sizes) based on the specific document type and detected tampering patterns. This dynamic parameter adjustment allows the double blurring correlation method to maintain high reliability across different document formats and tampering scenarios
3Measurement precision
If LBP-based techniques are used, then copy-paste and insertion tampering can be detected, but sophisticated tampering (e.g., image retouching) is missed
Solution Approach 1:
The detection system segments the overall tampering detection task into multiple specialized sub-tasks handled by different algorithms. LBP-based detection focuses on copy-paste and insertion patterns, while SVD and double blurring correlation handle sophisticated retouching and blending. This segmentation allows each algorithm to optimize for its specific detection domain while collectively covering all tampering types
Data Source
AI summary
Systems, apparatuses, methods, and computer program products are disclosed for detecting evidence of tampering in a digital document. An example method includes receiving by communications hardware, the digital document and determining, by tampering detection circuitry, a tampered region classification result for a region of the digital document. The example method further includes in an instance in which the tampered region classification result indicates tampering, providing, by the tampering detection circuitry, an indication of the region of the digital document and the tampered region classification result to a combination model and receiving, by the tampering detection circuitry, an overall tampering probability from the combination model.