On Constructing A Better Correlation Predictor For Prnu-Based Image Forgery Localization

The advances in image editing and retouching technology have enabled an unskilled person to easily produce visually indistinguishable forged images. To detect and localize such forgeries, many image forensic tools rely on visually imperceptible clues, e.g. the subtle traces or artifacts introduced during image acquisition and processing, and slide a regular, typically square, detection window across the image to search for discrepancies of specific clues. Such a sliding-window paradigm confines the explorable neighborhood to a regular grid and inevitably limits its capability in localizing forgeries in a broad range of shapes. While image segmentation that generates segments adhering to object boundaries might be a promising alternative to the sliding window-based approach, it is generally believed that the potential of the segmentation-based detection scheme is hindered by object removal forgeries where meaningful segments are often unavailable. In this work, we take forgery localization based on photo-response non-uniformity (PRNU) noise as an example and propose a segmentation-based forgery localization scheme that exploits the local homogeneity of visually indiscernible clues to mitigate the limitations of existing segmentation approaches that are merely based on visually perceptible content. We further propose a multi-orientation localization scheme that integrates the forgery probabilities obtained with image segmentation and multi-orientated detection windows. The multi-orientation scheme aggregates the complementary strengths of image segmentation and multi-oriented detection window in localizing the <italic>object insert</italic> and <italic>object removal</italic> forgeries. Experimental results on a public realistic tampering image dataset demonstrate that the proposed segmentation-based and multi-orientation forgery localization schemes outperform existing state-of-the-art PRNU-based forgery localizers in terms of both region and boundary <inline-formula> <tex-math notation="LaTeX">$F_{1}$ </tex-math></inline-formula> scores.

This paper presents an algorithm to solve the problem of Photo-Response Non-Uniformity (PRNU) noise facing stabilized video. The stabilized video undergoes in-camera processing like rolling shutter correction. Thus, misalignment exists between the PRNU noises in the adjacent frames owing to the global and local frame registration performed by the in-camera processing. The misalignment makes the reference PRNU noise and the test PRNU noise unable to extract and match accurately. We design a computing method of maximum likelihood estimation algorithm for extracting the PRNU noise from stabilized video frames. Besides, unlike most prior arts tending to match the PRNU noise in whole frame, we propose a new patch-based matching strategy, aiming at reducing the influence from misalignment of frame the PRNU noise. After extracting the reference PRNU noise and the test PRNU noise, this paper adopts the reference and the test PRNU overlapping patch-based matching. It is different from the traditional matching method. This paper conducts different experiments on 224 stabilized videos taken by 13 smartphones in the VISION database. The area under curve of the algorithm proposed in this paper is 0.841, which is significantly higher than 0.805 of the whole frame matching in the traditional algorithm. Experimental results show good performance and effectiveness the proposed strategy by comparing with the prior arts.

Photo-response non-uniformity (PRNU) noise has been well established as a reliable fingerprint of imaging sensors for source camera identification and other forensic applications. In this paper, we introduce a novel denoising method for PRNU noise extraction that considerably outperforms the existing techniques. The rationale is to use as little as one adjacent pixel in spatial domain filtering to suppress the pixel-to-pixel correlation in the estimation of the (supposedly white) PRNU noise. Experimental results are presented based on an image database of ten cameras of different models and makes. These results are embodied in the receiver operating characteristic (ROC) of source camera identification.

The increasing prevalence of digital technology brings great convenience to human life, while also shows us the problems and challenges. Relying on easy-to-use image editing tools, some malicious manipulations, such as image forgery, have already threatened the authenticity of information, especially the electronic evidence in the crimes. As a result, digital forensics attracts more and more attention of researchers. Since some general post-operations, like widely used smooth filtering, can affect the reliability of forensic methods in various ways, it is also significant to detect them. Furthermore, the determination of detailed filtering parameters assists to recover the tampering history of an image. To deal with this problem, we propose a new approach based on convolutional neural networks (CNNs). Through adding a transform layer, obtained distinguishable frequency-domain features are put into a conventional CNN model, to identify the template parameters of various types of spatial smooth filtering operations, such as average, Gaussian and median filtering. Experimental results on a composite database show that putting the images directly into the conventional CNN model without transformation can not work well, and our method achieves better performance than some other applicable related methods, especially in the scenarios of small size and JPEG compression.

The ever increasing prevalence of smartphones and the popularity of social network platforms have facilitated instant sharing of multimedia content through social networks. However, the ease in taking and sharing photos and videos through social networks also allows privacy-intrusive and illegal content to be widely distributed. As such, images captured and shared by users on their profiles are considered as significant digital evidence for social network data analysis. The Sensor Pattern Noise (SPN) caused by camera sensor imperfections during the manufacturing process mainly consists of the Photo-Response Non-Uniformity (PRNU) noise that can be extracted from taken images without hacking the device. It has been proven to be an effective and robust device fingerprint that can be used for different important digital image forensic tasks, such as image forgery detection, source device identification and device linking. Particularly, by fingerprinting the camera sources captured a set of shared images on social networks, User Profile Linking (UPL) can be performed on social network platforms. The aim of this thesis is to present effective and robust methods and algorithms for better fulfilling shared image analysis based on SPN. We propose clustering and classification based methods to achieve Smartphone Identification (SI) and UPL tasks, given a set of images captured by a known number of smartphones and shared on a set of known user profiles. The important outcome of the proposed methods is UPL across different social networks where the clustered images from one social network are applied to fingerprint the related smartphones and link user profiles on the other social network. Also, we propose two methods for large-scale image clustering of different types of the shared images by users, without prior knowledge about the types and number of the smartphones.

Capturing images has been increasingly popular in recent years, owing to the widespread availability of cameras. Images are essential in our daily lives because they contain a wealth of information, and it is often required to enhance images to obtain additional information. A variety of tools are available to improve image quality; nevertheless, they are also frequently used to falsify images, resulting in the spread of misinformation. This increases the severity and frequency of image forgeries, which is now a major source of concern. Numerous traditional techniques have been developed over time to detect image forgeries. In recent years, convolutional neural networks (CNNs) have received much attention, and CNN has also influenced the field of image forgery detection. However, most image forgery techniques based on CNN that exist in the literature are limited to detecting a specific type of forgery (either image splicing or copy-move). As a result, a technique capable of efficiently and accurately detecting the presence of unseen forgeries in an image is required. In this paper, we introduce a robust deep learning based system for identifying image forgeries in the context of double image compression. The difference between an image’s original and recompressed versions is used to train our model. The proposed model is lightweight, and its performance demonstrates that it is faster than state-of-the-art approaches. The experiment results are encouraging, with an overall validation accuracy of 92.23%.

The key to preventing the COVID-19 is to diagnose patients quickly and accurately. Studies have shown that using Convolutional Neural Networks (CNN) to analyze chest Computed Tomography (CT) images is helpful for timely COVID-19 diagnosis. However, personal privacy issues, public chest CT data sets are relatively few, which has limited CNN's application to COVID-19 diagnosis. Also, many CNNs have complex structures and massive parameters. Even if equipped with the dedicated Graphics Processing Unit (GPU) for acceleration, it still takes a long time, which is not conductive to widespread application. To solve above problems, this paper proposes a lightweight CNN classification model based on transfer learning. Use the lightweight CNN MobileNetV2 as the backbone of the model to solve the shortage of hardware resources and computing power. In order to alleviate the problem of model overfitting caused by insufficient data set, transfer learning is used to train the model. The study first exploits the weight parameters trained on the ImageNet database to initialize the MobileNetV2 network, and then retrain the model based on the CT image data set provided by Kaggle. Experimental results on a computer equipped only with the Central Processing Unit (CPU) show that it consumes only 1.06 s on average to diagnose a chest CT image. Compared to other lightweight models, the proposed model has a higher classification accuracy and reliability while having a lightweight architecture and few parameters, which can be easily applied to computers without GPU acceleration. Code:github.com/ZhouJie-520/paper-codes.

Image splicing detection based on convolutional neural network with weight combination strategy

JPEG compression is one of image degradations that often occurs in image storing and retouching process. Estimating JPEG compression degradation property is important for JPEG deblocking algorithm and image forensic analysis. JPEG degradation exists not only in JPEG file format but also in other image formats because JPEG distortion remains after converting to another image format. Moreover, JPEG degradation property is not always uniform within an image in case that the image is collaged from different JPEG-compressed photos. In this paper, pixelwise detection of JPEG-compression degradation and estimation of JPEG quality factor using a convolutional neural network is proposed. The proposed network outputs an estimated JPEG quality factor map and a compression flag map from an input image. Experimental results show that the proposed network successfully infers the quality factors and discriminates between non-JPEG-compressed images and JPEG-compressed images. We also demonstrate that the proposed network can spot a collaged region in a fake image which is comprised of images that have different JPEG compression properties. Additionally, the network reveals that image datasets Set5 and Set14, often used to evaluate super-resolution algorithms, contain JPEG-compressed low quality images, which are inappropriate for such evaluation.

The photo-response non-uniformity (PRNU) noise of imaging sensors significantly aids digital forensics and judicial identification, as it can be used as the fingerprint for uniquely identifying individual imaging devices. During the PRNU fingerprint extraction, it is very important for source camera identification to estimate the natural noise from real-world images. Most existing deep learning-based neural networks have a large number of model parameters, and they may not be practical in realistic scenarios such as deploying the model on small devices like smartphones and remote forensics equipment. In this paper, we present a lightweight PRNU fingerprint extraction algorithm based on an invertible denoising network (InvDN) for source camera identification. Specifically, to reduce the number of parameters, the deep network uses a constant amount of memory to compute the gradient and employs the same parameters for both forward and backward propagation. In addition, by introducing an information-loss-less reversible network, more complete PRNU fingerprint information can be extracted. Experimental results show that this algorithm exhibits superior PRNU fingerprint extraction performance and generalization ability compared to the state-of-the-art PRNU fingerprint extraction algorithms.

Current prominent camera identification methods use wavelet-based filter to extract photo-response non-uniformity (PRNU) noise as camera fingerprint [1, 2]. However, these noise features in heavily textured images can not be extracted by using wavelet-based filter effectively. In this paper, we propose a new camera identification method that uses curvelet-based filter to extract noise features in heavily textured images or non-heavily textured image. Because curvelet transform allows an optimal sparse representation of objects with C2 singularities, curvelet-based filter can extract the noise features in heavily textured images more effectively than wavelet-based filter. To increase the recognition rate for heavily textured images, we differentiate heavily textured images from non-heavily textured images by using the bivariate kurtosis of an image, and Neyman-Pearson decision is used to determine different decision thresholds.

and AI-assisted image manipulation tools, which The emergence of advanced image manipulation techniques in the digital age has made it extremely difficult to detect and categorize forged images. In order to identify and classify image forgeries, this study proposes a deep learning-based method that combines a Convolutional Neural Network (CNN) classifier with the You Only Look Once (YOLO) object detection model. The suggested system uses sophisticated feature extraction techniques to categorize images into authentic, copy-move, spliced, and deepfake forgeries after being trained on a custom dataset of manipulated images. While CNN-based classification offers accurate categorization, our experimental results show how reliable and effective YOLO is at locating forgery regions. Because of its high accuracy, the suggested model is a useful tool for security and digital forensics applications. Keywords: Image Forgery, YOLO, CNN, Deep Learning, Classification, Digital Forensics

One of the most popular image forgery technique is copy-move forgery. In this technique, one or more segments are copied and affixed at different positions within the image. This forgery technique is highly grievous as it can manipulate an image in various ways (such as by presenting additional information or by concealing the genuine information of image). We propose a novel blind forensic technique for copy-move image forgery detection. Our approach utilize different edge detection operators to extract high frequency features. Histogram of Oriented Gradients (HOG) and Weber Local Descriptor (WLD) are used to extract image block features. Radix and lexicographical sorting is enforced over feature vector matrix followed by correlation computation between feature vectors to detect similar feature vectors. Shift vectors are computed to locate similar group of blocks within image. Connected component labeling is applied as morphological operation to remove false matches. Proposed approach is robust to detect plain as well as multiple copy-move forgery in images with post-processing attacks such as contrast adjustment, image blurring, color reduction, and brightness change. Proposed approach achieve highest F-Measure(%) in comparision to other existing forgery detection methods.

To extra validate and evaluate the reproducibility of a commercial deep convolutional neural network (DCNN) algorithm for pulmonary nodules on chest radiographs (CRs) and to compare its performance with radiologists. This retrospective study enrolled 434 CRs (normal to abnormal ratio, 246:188) from 378 patients that visited a tertiary hospital. DCNN performance was compared with two radiology residents and two thoracic radiologists. Abnormality assessment (using the area under the receiver operating characteristics (AUROC)) and nodule detection (using jackknife alternative free-response ROC (JAFROC)) were compared among three groups (DCNN only, radiologist without DCNN and radiologist with DCNN). A subset of 56 paired cases, having two CRs taken within a 7-day period, were assessed for intraobserver reproducibility using the intraclass correlation coefficient. Independent characteristics of pulmonary nodules detected by DCNN were assessed by multiple logistic regression analysis. The AUROC for abnormality detection for the three groups were 0.87, 0.93 and 0.96, respectively (P<0.05), whereas the JAFROC analysis of nodule detection was 0.926, 0.929 and 0.964. Reproducibility for the three groups was 0.80, 0.67 and 0.80, which shows an increase in radiologists using DCNN (P<0.05). Nodules detected by DCNN were more solid, round-shaped and well marginated, not masked and laterally located (P<0.05). Extra validation results of DCNN showed high ROC results and there was a significant improvement in the performance when radiologists used DCNN. Reproducibility by DCNN alone showed good agreement, and there was an improvement from moderate to good agreement for radiologists using DCNN.

Recent advances in deep learning have significantly propelled the development of image forgery localization. However, existing models remain highly vulnerable to adversarial attacks: imperceptible noise added to forged images can severely mislead these models. In this paper, we address this challenge with an Adversarial Noise Suppression Module (ANSM) that generates a defensive perturbation to suppress the attack effect of adversarial noise. We observe that forgery-relevant features extracted from adversarial and original forged images exhibit distinct distributions. To bridge this gap, we introduce Forgery-relevant Features Alignment (FFA) as a first-stage training strategy, which reduces distributional discrepancies by minimizing the channel-wise Kullback-Leibler divergence between these features. To further refine the defensive perturbation, we design a second-stage training strategy, termed Mask-guided Refinement (MgR), which incorporates a dual-mask constraint. MgR ensures that the defensive perturbation remains effective for both adversarial and original forged images, recovering forgery localization accuracy to their original level. Extensive experiments across various attack algorithms demonstrate that our method significantly restores the forgery localization model's performance on adversarial images. Notably, when ANSM is applied to original forged images, the performance remains nearly unaffected. To our best knowledge, this is the first report of adversarial defense in image forgery localization tasks.