Description:FIELD OF THE INVENTION

[0001] The present invention relates to a system for detecting and localizing tampered regions in digital images that is capable of enabling robust detection and localization of tampered regions across diverse image formats and manipulation techniques.

BACKGROUND OF THE INVENTION

[0002] The rapid advancements in digital image editing tools have made tampering a significant concern in domains like legal evidence validation, media authentication, and cybersecurity. Existing methods, primarily based on convolutional neural networks (CNNs), struggle with capturing long-range dependencies and ensuring computational efficiency, especially for complex image manipulations such as splicing, copy-move, and inpainting. Transformer models offer enhanced global contextual understanding but are computationally intensive, limiting their applicability in resource-constrained environments. This gap underscores the need for an efficient, hybrid approach that balances precision and scalability. The proposed EffiUNet Transformer framework addresses these challenges by integrating the structural efficiency of U-Net with transformer-based arrangements, enabling robust detection and localization of tampered regions across diverse image formats and manipulation techniques. This solution aims to enhance forensic image analysis by offering high accuracy, adaptability, and real-time deployment capabilities

[0003] Digital image tampering detection has been an area of extensive research due to its critical applications in forensics, media integrity, and cybersecurity. Traditional methods often relied on handcrafted features and heuristic approaches, which proved inadequate in detecting sophisticated manipulations like splicing and deepfake generation. Modern solutions have predominantly adopted convolutional neural networks (CNNs) for their ability to learn feature representations from data. However, CNNs face limitations in capturing global contextual relationships and detecting subtle inconsistencies in manipulated regions.

[0004] US9876543B2 discloses a system and method to provide improved channel estimation and channel state information (CSI) transfer in a wireless data communication system, including but not limited to Multiple Input Multiple Output (MIMO) communication systems. In accordance with one or more embodiments and aspects thereof, a channel estimation and CSI transfer system is disclosed that extends and utilizes existing frame transfer elements to accomplish estimation and transfer. Such a system may offer improved capabilities such as a lower overhead CSI transfer, reduced data transfer latency, and more frequent CSI measurements.

[0005] EP2345678A1 Discloses a recording material including a certain support and disposed thereon at least one layer including certain core/shell polymeric particles, the particles having, when dry, at least one void is provided. A method for providing an image using the recording sheet is also provided.

[0006] Conventionally, many systems have been developed to detect and localize tempered region in digital images they often lack robustness across diverse manipulation techniques and fail to provide pixel-level localization. Additionally, high computational costs make these tools impractical for real-time or resource-constrained applications.

[0007] In order to overcome the aforementioned drawbacks, there exists a need in the art to develop a system that is capable of focusing more on detection than localization, limiting their utility in forensic scenarios requiring precise tampering boundary identification and scalability for diverse datasets.

OBJECTS OF THE INVENTION

[0008] The principal object of the present invention is to overcome the disadvantages of the prior art.

[0009] An object of the present invention is to develop a system that is capable of identifying and highlighting tampered areas with high accuracy by capturing both local details and global context, enabling precise tampering detection across various image formats.

[0010] Another object of the present invention is to develop a system that is capable of enhancing the feature processing for extracting image features at different resolutions to detect both small and large tampered regions, ensuring comprehensive analysis of diverse tampering.

[0011] Another object of the present invention is to develop a system that is capable of analyzing the entire image to detect subtle tampering patterns, such as texture inconsistencies or edge irregularities, by understanding global relationships in the image data.

[0012] Yet another object of the present invention is to develop a system that is capable of reducing computational requirements to optimize the model to run quickly on devices like mobile phones or embedded systems, making real-time tamper detection practical for forensic and cybersecurity applications.

[0013] The foregoing and other objects, features, and advantages of the present invention will become readily apparent upon further review of the following detailed description of the preferred embodiment as illustrated in the accompanying drawings.

SUMMARY OF THE INVENTION

[0014] The present invention relates to a system for detecting and localizing tampered regions in digital images that is capable of focusing on detection than localization, limiting their utility in forensic scenarios and reducing computational requirements to optimize the model to run quickly on devices like mobile phones or embedded systems, making real-time tamper detection practical for forensic and cybersecurity applications.

[0015] According to an embodiment of the present invention, a system for detecting and localizing tampered regions in digital images comprising of a EffiUNet transformer framework module that combines a lightweight U-net architecture with transformer-based attention modules to process digital images to identify and highlight tampered areas with high accuracy by capturing both local details and global context, the EffiUNet Transformer Framework Module further includes a data augmentation module to add an artificial tampering effects like rotation, scaling, and noise to training images, improving the model’s ability to detect diverse and complex tampering techniques, a feature extraction module integrated into the encoder of the EffiUNet transformer framework module, connecting to transformer blocks for enhanced feature processing for extracting image features at different resolutions to detect both small and large tampered regions, the multi-scale feature extractor module is enhanced by an adversarial training, module and a transformer attention module embedded within the EffiUNet framework module, linked between the encoder and decoder paths of the U-Net structure to analyzes the entire image to detect subtle tampering patterns, such as texture inconsistencies or edge irregularities, by understanding global relationships in the image data.

[0016] According to another embodiment of the present invention, the system further comprises of a tamper localization module integrated into the decoder of the EffiUNet transformer framework module for receiving processed features from the contextual transformer block, in order to produce a detailed map highlighting tampered regions in the image, allowing forensic analysts to pinpoint manipulated areas with pixel-level accuracy, the segmentation map generator module is supported by a visualization interface module, connected to the output of the segmentation map generator for post-processing to displays tampered regions in a user-friendly format, enabling forensic analysts to easily interpret and use the localization results in investigations, a lightweight optimization module applied across the EffiUNet transformer framework module to reduce computational requirements to optimize the model to run quickly on devices like mobile phones or embedded systems, making real-time tamper detection practical for forensic and cybersecurity applications and the computational efficiency optimizer Module incorporates a modular integration module, linked to the EffiUNet framework module for compatibility with external systems.

[0017] While the invention has been described and shown with particular reference to the preferred embodiment, it will be apparent that variations might be possible that would fall within the scope of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

[0018] These and other features, aspects, and advantages of the present invention will become better understood with regard to the following description, appended claims, and accompanying drawings where:
Figure 1 illustrates flowchart depicting a system for detecting and localizing tampered regions in digital images

DETAILED DESCRIPTION OF THE INVENTION

[0019] The following description includes the preferred best mode of one embodiment of the present invention. It will be clear from this description of the invention that the invention is not limited to these illustrated embodiments but that the invention also includes a variety of modifications and embodiments thereto. Therefore, the present description should be seen as illustrative and not limiting. While the invention is susceptible to various modifications and alternative constructions, it should be understood, that there is no intention to limit the invention to the specific form disclosed, but, on the contrary, the invention is to cover all modifications, alternative constructions, and equivalents falling within the spirit and scope of the invention as defined in the claims.

[0020] In any embodiment described herein, the open-ended terms "comprising," "comprises,” and the like (which are synonymous with "including," "having” and "characterized by") may be replaced by the respective partially closed phrases "consisting essentially of," consists essentially of," and the like or the respective closed phrases "consisting of," "consists of, the like.

[0021] As used herein, the singular forms “a,” “an,” and “the” designate both the singular and the plural, unless expressly stated to designate the singular only.

[0022] The present invention relates to a system for detecting and localizing tampered regions in digital images that is capable of identifying and highlighting tampered areas with high accuracy by capturing both local details and global context and allowing to produce a detailed map highlighting tampered regions in the image, allowing forensic analysts to pinpoint manipulated areas with pixel-level accuracy.

[0023] Referring to Figure 1, a flowchart depicting a system for detecting and localizing tampered regions in digital images. The system discloses herein includes a EffiUNet transformer framework module that combines a lightweight U-net architecture with transformer-based attention modules to processe digital images to identify and highlight tampered areas with high accuracy by capturing both local details and global context, enabling precise tampering detection across various image formats. The EffiUNet transformer framework integrates a lightweight U-Net architecture with transformer-based attention modules to enhance tampering detection by effectively capturing both local and global image features. The U-Net component employs encoder-decoder pathways with skip connections to extract fine-grained spatial details, ensuring precise localization of tampered regions. Embedded within this structure, transformer modules utilize self-attention to model long-range dependencies and contextual relationships across the entire image, enabling the network to identify subtle inconsistencies indicative of manipulation. This combination allows the framework to process digital images efficiently maintaining computational lightweightness while achieving high accuracy in detecting tampered areas across various formats. The model learns to highlight anomalies by leveraging local feature extraction from the U-Net and global contextual understanding from the transformer, resulting in robust and precise tampering segmentation even in complex or diverse image scenarios.

[0024] The EffiUNet Transformer Framework Module further includes a data augmentation module to add an artificial tampering effects like rotation, scaling, and noise to training images, improving the model’s ability to detect diverse and complex tampering techniques. The EffiUNet Transformer Framework incorporates a data augmentation module that systematically applies artificial tampering transformations such as rotation, scaling, noise addition, and other manipulations to the training images, thereby enriching the dataset with a wide variety of altered samples. This process involves real-time, random application of these transformations during training, which simulates diverse tampering scenarios and enhances the model’s robustness. By exposing the network to a broad spectrum of artificially manipulated images, the augmentation module enables the EffiUNet to learn more generalized features of tampering, improving its capacity to detect subtle and complex manipulations across different formats and real-world conditions. The augmented data effectively trains the model to recognize a wide array of tampering artifacts, leading to greater accuracy and resilience against unforeseen or sophisticated tampering techniques during inference.

[0025] A feature extraction module integrated into the encoder of the EffiUNet transformer framework module, connecting to transformer blocks for enhanced feature processing for extracting image features at different resolutions to detect both small and large tampered regions, ensuring comprehensive analysis of diverse tampering types like splicing and imprinting. The feature extraction module integrated into the encoder of the EffiUNet transformer framework operates by employing multi-scale convolutional layers such as residual blocks or dilated convolutions that capture rich hierarchical features at various resolutions, effectively encoding both fine details and coarse contextual information. These extracted features are progressively down sampled to generate multi-resolution feature maps, which are then fed into transformer blocks equipped with self-attention. The transformer blocks further enhance these features by modeling long-range dependencies and global context across different spatial scales, enabling the network to identify subtle anomalies and large tampered regions alike. This combined approach allows the framework to detect diverse tampering types such as splicing (which may involve localized manipulations) and imprinting (often affecting larger areas) by ensuring comprehensive analysis at multiple resolutions, thus improving the model’s ability to accurately localize and characterize tampered regions of varying sizes and complexities.

[0026] The multi-scale feature extractor module is enhanced by an adversarial training, module. The multi-scale feature extractor through a process where a discriminator network is trained alongside the feature extractor to distinguish between real and manipulated feature representations. During training, the feature extractor generates multi-scale features from input images, which are then evaluated by the discriminator to assess their authenticity or consistency with genuine features. The feature extractor aims to produce representations that are indistinguishable from real, untampered features, while the discriminator learns to improve its ability to detect subtle discrepancies. This adversarial dynamic encourages the feature extractor to learn more discriminative, robust, and high-fidelity multi-scale features that better capture subtle tampering cues across different resolutions. Consequently, the enhanced features improve the model’s ability to detect both obvious and covert tampering artifacts, leading to more accurate and resilient tampering localization and classification.

[0027] A transformer attention module embedded within the EffiUNet framework module, linked between the encoder and decoder paths of the U-Net structure to analyzes the entire image to detect subtle tampering patterns, such as texture inconsistencies or edge irregularities, by understanding global relationships in the image data. Within the EffiUNet framework, the transformer attention module functioning as a bridge that enhances feature representations with global context. After the encoder extracts multi-scale features, the attention module applies self-attention to these features, allowing each spatial position to dynamically weigh its relevance to all other positions across the entire image. This process captures long-range dependencies and subtle contextual cues, such as texture inconsistencies or edge irregularities indicative of tampering. By modeling the global relationships, the transformer attention module integrates detailed local features with overarching spatial correlations, enabling the decoder to more effectively identify subtle manipulation artifacts that dispersed or masked within the image, thus improving the detection of nuanced tampering patterns across different regions and scales.

[0028] The segmentation map generator module is supported by a visualization interface module, connected to the output of the segmentation map generator for post-processing to displays tampered regions in a user-friendly format, enabling forensic analysts to easily interpret and use the localization results in investigations. The segmentation map generator module produces pixel-wise predictions that delineate tampered regions within the input image by leveraging learned features and spatial context, resulting in a binary or multi-class map highlighting manipulated areas. This map is then fed into the visualization interface module, which acts as a post-processing layer that overlays the segmentation results onto the original image, often with color coding or transparency to clearly distinguish tampered from authentic regions. The interface include tools for zooming, annotation, or exporting, enabling forensic analysts to easily interpret the localized manipulations visually. By translating complex model outputs into an intuitive, user-friendly format, the visualization interface facilitates rapid assessment and decision-making in investigative scenarios, making the detection results accessible and actionable in real-world forensic workflows.

[0029] A tamper localization module integrated into the decoder of the EffiUNet transformer framework module for receiving processed features from the contextual transformer block, in order to produce a detailed map highlighting tampered regions in the image, allowing forensic analysts to pinpoint manipulated areas with pixel-level accuracy. The tamper localization module integrated into the EffiUNet decoder receives refined, context-rich features from the transformer-based contextual module, which have been enhanced with global relationships and subtle pattern information. This module applies a series of convolutional layers, often combined with up sampling operations, to progressively refine these features and generate a dense, pixel-level probability map that indicates the likelihood of tampering at each location. By leveraging skip connections from earlier encoder layers, it preserves fine-grained details necessary for precise localization. The final output is a detailed tamper map, processed further, that highlights manipulated regions with pixel-level accuracy, enabling forensic analysts to precisely identify and examine the manipulated areas within the image for investigative purposes.

[0030] A lightweight optimization module applied across the EffiUNet transformer framework module to reduce computational requirements to optimize the model to run quickly on devices like mobile phones or embedded systems, making real-time tamper detection practical for forensic and cybersecurity applications. The lightweight optimization module applied across the EffiUNet transformer framework employs techniques such as model pruning, quantization, and efficient convolutional operations to reduce computational complexity and memory footprint. For instance, it might prune redundant or less important weights to streamline the network, quantize weights and activations to lower precision and replace standard convolutions with depthwise convolutions that maintain accuracy while decreasing computations. Additionally, it optimizes the transformer components by limiting the number of attention heads or embedding dimensions, and employs efficient attention arrangements designed for low-resource environments. These combined strategies enable the model to perform fast, real-time tamper detection on resource-constrained devices like mobile phones and embedded systems, making it practical for on-the-spot forensic and cybersecurity tasks without sacrificing critical accuracy.

[0031] The computational efficiency optimizer Module incorporates a modular integration module, linked to the EffiUNet framework module for compatibility with external systems. The Computational Efficiency Optimizer Module functions by integrating a modular compatibility interface within the EffiUNet framework, enabling seamless communication and interoperability with external systems such as cybersecurity platforms or forensic analysis tools. It achieves this by standardizing input/output formats, employing adaptable API endpoints, and incorporating lightweight, flexible connectors that facilitate data exchange without extensive reconfiguration. Technically, this module encapsulates optimized data preprocessing, model inference, and post-processing routines into standardized, modular components that easily plugged into external workflows. It also includes resource-aware scheduling and adaptive load balancing features, ensuring efficient operation across diverse environments. By providing a unified, modular interface, it ensures the EffiUNet framework integrated into broader forensic or cybersecurity systems, allowing real-time tamper detection results to be shared, visualized, or further analyzed with minimal latency and maximal compatibility.

[0032] The present invention works best in the following manner, the EffiUNet transformer framework module that combines a lightweight U-net architecture with transformer-based attention modules to processe digital images to identify and highlight tampered areas with high accuracy by capturing both local details and global context. The EffiUNet Transformer Framework Module further includes a data augmentation module to add an artificial tampering effects like rotation, scaling, and noise to training images, improving the model’s ability to detect diverse and complex tampering techniques. The feature extraction connecting to transformer blocks for enhanced feature processing for extracting image features at different resolutions to detect both small and large tampered regions, ensuring comprehensive analysis of diverse tampering types like splicing and inpainting. The multi-scale feature extractor module is enhanced by an adversarial training, module The transformer attention module embedded within the EffiUNet framework module, linked between the encoder and decoder paths of the U-Net structure to analyzes the entire image to detect subtle tampering patterns, such as texture inconsistencies or edge irregularities, by understanding global relationships in the image data. The segmentation map generator module is supported by a visualization interface module, connected to the output of the segmentation map generator for post-processing to displays tampered regions in a user-friendly format, enabling forensic analysts to easily interpret and use the localization results in investigations. The lightweight optimization module applied across the EffiUNet transformer framework module to reduce computational requirements to optimize the model to run quickly on devices like mobile phones or embedded systems, making real-time tamper detection practical for forensic and cybersecurity applications. The computational efficiency optimizer Module incorporates a modular integration module, linked to the EffiUNet framework module for compatibility with external systems.

[0033] Although the field of the invention has been described herein with limited reference to specific embodiments, this description is not meant to be construed in a limiting sense. Various modifications of the disclosed embodiments, as well as alternate embodiments of the invention, will become apparent to persons skilled in the art upon reference to the description of the invention. , Claims:1) A system for detecting and localizing tampered regions in digital images, comprising:

i) a EffiUNet transformer framework module that combines a lightweight U-net architecture with transformer-based attention modules to processe digital images to identify and highlight tampered areas with high accuracy by capturing both local details and global context, enabling precise tampering detection across various image formats;
ii) a feature extraction module integrated into the encoder of the EffiUNet transformer framework module, connecting to transformer blocks for enhanced feature processing for extracting image features at different resolutions to detect both small and large tampered regions, ensuring comprehensive analysis of diverse tampering types like splicing and inpainting;
iii) a transformer attention module embedded within the EffiUNet framework module, linked between the encoder and decoder paths of the U-Net structure to analyzes the entire image to detect subtle tampering patterns, such as texture inconsistencies or edge irregularities, by understanding global relationships in the image data;
iv) a tamper localization module integrated into the decoder of the EffiUNet transformer framework module for receiving processed features from the contextual transformer block, in order to produce a detailed map highlighting tampered regions in the image, allowing forensic analysts to pinpoint manipulated areas with pixel-level accuracy; and
v) a lightweight optimization module applied across the EffiUNet transformer framework module to reduce computational requirements to optimize the model to run quickly on devices like mobile phones or embedded systems, making real-time tamper detection practical for forensic and cybersecurity applications.

2) The system as claimed in claim 1, wherein the EffiUNet Transformer Framework Module further includes a data augmentation module to add an artificial tampering effects like rotation, scaling, and noise to training images, improving the model’s ability to detect diverse and complex tampering techniques.

3) The system as claimed in claim 1, wherein the multi-scale feature extractor module is enhanced by an adversarial training, module.

4) The system as claimed in claim 1, wherein the segmentation map generator module is supported by a visualization interface module, connected to the output of the segmentation map generator for post-processing to displays tampered regions in a user-friendly format, enabling forensic analysts to easily interpret and use the localization results in investigations.

5) The system as claimed in claim 1, wherein the computational efficiency optimizer Module incorporates a modular integration module, linked to the EffiUNet framework module for compatibility with external systems.