Camera model identification remains a core challenge in digital image forensics for source attribution, integrity verification, and copyright protection. While rich model residuals effectively capture sensor and in-camera processing artifacts, most existing approaches rely on high-dimensional residual sets, leading to redundancy, high computational cost, and limited exploration of large-scale multi-class settings. This work introduces a systematic feature-engineering framework that selectively retains only the most discriminative residual components, reducing dimensionality from 34,671 to 7,176 features. This approach significantly enhances computational efficiency and class separability while preserving forensic performance. The method is explicitly optimized for scalable multi-class camera model identification, providing a lightweight yet robust alternative to both traditional and deep-learning-based techniques. Extensive experiments were conducted on the Dresden and VISION datasets across inter-brand, intra-brand, and large-scale multi-class scenarios. On 18 and 23 camera models from Dresden and 29 camera models from VISION, the proposed framework achieved classification accuracies of 98.22%, 97.12% and 96.45%, respectively, demonstrating strong scalability and generalization. The robustness was further validated under realistic post-processing transformations, including JPEG compression (QF = 50 and QF = 80), resizing, and center cropping. The model maintained high diagonal dominance in confusion matrices under these degradations, confirming resilience to common real-world image manipulations. Overall, the selective residual modeling combined with non-linear kernel-based classification delivers a novel, computationally efficient, and high-performing solution for large-scale camera model identification. It outperforms or matches recent deep learning approaches while requiring significantly fewer computational resources, making it particularly suitable for practical forensic applications.

As images are integral to many sectors in the digital age, it is essential to ensure their authenticity and integrity. However, the ease of digital image creation and sharing also exposes them to manipulation and misrepresentation, heightening concerns about privacy and misinformation. The process of recapturing, a prevalent anti-forensic technique, poses challenges to tampering detection methods, necessitating effective countermeasures to uphold image credibility. Monitor-screenshots, facilitated by the simplicity of capturing screenshots of Original images displayed on monitors, pose unique challenges in source identification. Addressing this, we propose a novel approach to unveil the screen fingerprint, capturing distinctive irregularities associated with blur exist in Monitor-screenshots for accurate source identification. Leveraging image registration, difference image masking, and sophisticated feature extraction techniques, our method enables precise identification of specific screens, enhancing the authentication of digital content. By scrutinizing screen-specific characteristics and artifacts left during recapture, proposed model can verify the claimed origin of Screenshots, tested on a Screenshot dataset using SVM classifier, offers a robust framework to authenticate digital content and trace its source with precision and reliability, mitigating risks associated with image manipulation and misrepresentation in the digital domain.

A decade of research has been conducted on detecting copy-move forgeries (CMFD). Technology has enabled the manipulation of images, once the most authentic source of information. This paper proposes a copy-move forgery detection algorithm based on fused features to address issues such as time complexity and difficulty detecting forgeries in smooth regions. To extract descriptive features, a low-contrast threshold was used in conjunction with three detection methods, including scale-invariant feature transform (SIF), speeded-up robust features (SURF), and accelerated KAZE (AKAZE). SURF and accelerated KAZE (AKAZE) are used in our keypoint-based CMFD technique. To detect manipulated regions efficiently, AKZAE, SURF, and SIFT can be used to extract major keypoints in smooth regions.

There have been numerous advancements in information exchange and covert information exchange in recent years by virtue of the various methodologies proposed. Thus, there is a need to keep up with the said advancements' framework. This paper reviews the progress of a covert information exchange technique, Steganography, over the recent few years by sifting through some cornerstone techniques and papers. We have listed and reviewed the various types of Steganography, with special emphasis on Image Steganography and their various techniques and methodologies. Through this paper, some of the methodologies proposed in recent years are analysed and their characteristics are discussed.

Diagrams are the most important part of any project. These diagrams help in better analyzing, understanding and optimizing the problem solution. UML diagram is one of the most important models to understand software modules and generally they are exported in various image formats. Images of these diagrams will lead to loss of their structural information. This structural file is used as a modeling tool for recognition, visualization and processing of the diagram. If we want to understand or check any vulnerability, ambiguity and error in the diagram, that has to be done manually and requires high skill and more time. This problem can be alleviated if we have such software that recognizes the engineering diagram from images and converts them in a format which is recognizable by their case tool software. Hand drawn graph recognition works for recognizing these images of engineering diagrams.

Image forensics researchers have recently focused a lot of attention on detection of median filtering that is used to hide evidence of image forgery and related operations. After median filtering, JPEG compression, Gaussian blur, noise addition, and resampling are often applied to further hide traces of median filtering. Researchers are essentially abandoning a well-developed and established field of feature engineering for deep learning-based system that do not require manual feature engineering. In this paper, we combined the feature engineering and machine learning to develop a solution for robust median filtering detection. For the purpose, we developed an 18-D feature based on textured features and a neural network called MedFCN to performing classification between original and post-median filtering modifications by operations such as Gaussian blur, resampling, the addition of additive white Gaussian noise (AWGN), and JPEG compression. All datasets used in the study were created from UCID, BOSS, RAISE, and Dresden for training and testing of the system. By designing a multilayer perceptron that can accept an engineered feature set, we were able to combine two methodologies, feature engineering and deep learning, to develop a system that outperforms state-of-the-art methods for median filter forensics.

Trust is an essential security component and plays a crucial role in building secure information processing systems. Many organizations have adopted zero-trust principles in their security mechanisms, especially in Cloud Computing environments, where everything is presumed to be semi-honest and untrustworthy until proven otherwise. This is in blunt contrast to the traditional perimeter-security model, where it is presumed that the malicious actors are always outside the perimeter of the trusted network, and trustworthy users are always inside the network. One of the most significant challenges with the notion of trust is that till now more focus was put on the subjective aspect of trust which was challenging to evaluate and verify. There is an urgent need to address the objective and verifiable notion of trust, especially in the systems involving Federated Learning as in Cloud environments. We propose a Rich model based zero-trust security framework for trust verification of SaaS and employ machine learning functionalities to perform multimedia data analytics over the service processing behaviors for improved visibility into service operations and risks. Our proposed model relies heavily on the features extracted form Federated learning on the data of different cloud service users that is subjected to AI processing. We utilized the Rich models for feature extraction of large scale multimedia data for monitoring cloud service behavior and employed Ensemble classifier for analyzing the data features and employed majority voting of individual learners for decision making. The experiments performed on the standard dataset proved that the proposed model successfully monitored and analyzed the behavior of cloud services for validating the legitimacy of trusted SaaS and detecting any trust violations.

Human Activity Recognition is the act of recognizing activities performed by humans in real-time. This can be done using video data or more advanced forms of data like- inertial, depth maps, or human skeletal joint trajectories. In this work, we perform human action recognition through skeletal joint tracking of the human body using a deep recurrent neural network. Our proposed method was then tested on two standard databases, namely UTD-MHAD and MSR- Daily Activity 3D-Datasets. The judgement on the efficiency of our proposed model was made by comparing it to various, recently published, State-Of-The-Art (SOTA) methods.The evaluations of our model show that our method performs well on both the datasets and achieves an accuracy of 99.07%, and 91%, on UTD-MHAD and MSR Daily Activity databases respectively, and can recognize human activities from a variety of domains.

When it comes to reducing impulsive noise from digital photos, the Median Filter (MF) is a nonlinear filter that can be employed effectively. This nonlinear filter is used to erase traces left by other linear filters due to the fact that they are nonlinear in nature. The application of a median filter on an image raises questions about the image’s genuineness when it is discovered. For the detection of median filtering, a slew of approaches has been developed. The main purpose of this paper is to explain different parts of median filter forensics, look at some new and existing techniques in median filter forensics, and compare the pros and cons of each technique. We also developed a taxonomy to broadly classify various methods proposed for median filter forensics. In addition, we also discussed and describe the popular testing procedure that researchers in the field are using to evaluate the median filtering detection methods and may be employed as a testing framework for future studies, for other operator detection as well as for general purpose image forensics.

Within digital forensics, a notable emphasis is placed on the detection of the application of fundamental image-editing operators, including but not limited to median filters, average filters, contrast enhancement, resampling, and various other operations closely associated with these techniques. When conducting a historical analysis of an image that has potentially undergone various modifications in the past, it is a logical initial approach to search for alterations made by fundamental operators. This paper presents the development of a deep-learning-based system designed for the purpose of detecting fundamental manipulation operations. The research involved training a multilayer perceptron using a feature set of 36 dimensions derived from the gray-level co-occurrence matrix, gray-level run-length matrix, and normalized streak area. The system detected median filtering, mean filtering, the introduction of additive white Gaussian noise, and the application of JPEG compression in digital Images. Our system, which utilizes a multilayer perceptron trained with a 36-feature set, achieved an accuracy of 99.46% and outperformed state-of-the-art deep-learning-based solutions, which achieved an accuracy of 97.89%.

Passive image forgery techniques are aimed at detecting image tampering without any a prior knowledge about the images. One of the major challenges in the field of image forensics is detection of basic operators such as application of median filtering. Median filter is one of the basic operators that is being used for malicious intent of hiding history of operations applied on an image. The studies related to detection of median filtering are becoming very popular in the field of digital image forensics. A large number of methods have been proposed for detection of median filtering. The paper presents improved percentage streak area based on streak effect in median filtered images. An ensemble classifier is used for detection of median filtered images by performing a multi class classification using bagged method with decision tree as weak learner. The results show that the classifier performs better than state-of-the-art methods for multiclass classification between original, median filtered and postJPEG compressed median filtered images.

Over the years, computer scientists are working on building models that aid the scientific community in many ways by cutting laboratory expenses or by saving time. Such models find useful applications in microscopy images as well. Determining the morphology of nanoparticles from Transmission Electron Microscopy images is a manual and cumbersome task. In the past years, scientists have tried to build models to automate the process of nanoparticle segmentation in microscopy images. This study focuses on finding the best segmentation model, which achieves high metrics and is robust to microscopy parameters. For this purpose, eight different models have been compared. The training dataset consists of 150 BF-TEM Platinum nanoparticle images containing 3629 nanoparticles of all kinds. Further, we examine the generalizability of the models on E-TEM Gold nanoparticle images. We also describe essential considerations while choosing a network for segmenting nanoparticle images that generalize well across the Platinum BF-TEM and Gold E-TEM nanoparticles dataset. The layer gradients are visualized to further explain the black-box nature of neural networks.

Over the past years, Cloud computing has become one of the most influential information technologies to combat computing needs because of its unprecedented advantages. In spite of all the social and economic benefits it provides, it has its own fair share of issues. These include privacy, security, virtualization, storage, and trust. The underlying issues of privacy, security, and trust are the major barriers to the adoption of cloud by individuals and organizations as a whole. Trust has been the least looked into since it includes both subjective and objective characteristics. There is a lack of review on trust models in this research domain. This paper focuses on getting insight into the nomenclature of trust, its classifications, trust dimensions and throws an insight into various trust models that exist in the current knowledge stack. Also, various trust evaluation measures are highlighted in this work. We also draw a comparative analysis of various trust evaluation models and metrics to better understand the notion of trust in cloud environments. Furthermore, this work brings into light some of the gaps and areas that need to be tackled toward solving the trust issues in cloud environments so as to provide a trustworthy cloud ecosystem. Lastly, we proposed a Machine Learning backed Rich model based solution for trust verification in Cloud Computing. We proposed an approach for verifying whether the right software is running for the correct services in a trusted manner by analyzing features generated from the output cloud processed data. The proposed scheme can be utilized for verifying the cloud trust in delivering services as expected that can be perceived as an initiative towards trust evaluation in cloud services employing Machine learning techniques. The experimental results prove that the proposed method verifies the service utilized with an accuracy of 99%.

Digital images are part of our lives but with the advancement of technology, the authenticity of images is in doubt. Image editing tools are used to tamper images and high-quality cameras are used to recapture tampered images to evade tamper detection. In general, tampering introduces artifacts in images and these artifacts are camouflaged by the re-acquisition process. The re-acquisition process makes forged image more like original which is hard to detect visually and statistically. Thus, existing forensic tools and techniques fail to detect tampering in reacquired or recaptured images. This paper proposes a novel technique to detect recaptured images by exploiting the high-level details present in images and based on that edge profile is obtained. Further, edges are classified into different groups. It has been observed that the number of edge pixels in these edge groups is different for original and recaptured images. Based on the number of pixels in edges, a feature vector is built and a system is trained using SVM classifier. The proposed method tested on two databases. The experimental results demonstrated that proposed method is better than existing techniques for recapture detection.

The most common phenomenon in anti-forensic approach is to conceal image tampering artifacts through recapturing. Hence, forensic algorithms fail to detect tampering in manipulated images. The easy way to produce a Recaptured image to cover the manipulations is capturing it from LCD screens. Despite hiding manipulations, Recaptured images itself contain some characteristics that can be exploited by various algorithms to identify Recaptured images. High-resolution images contain sucient information to analyze multiple characteristics, but identication of low-resolution Screen-Captured images is problematic. Since low-resolution images have enough information to extract valid features. In this paper, a Convolutional Neural Network (CNN) is proposed that not only able to classify high-resolution images but can also distinguish low-resolution images as Single-Captured or Screen-Captured nely. The network is trained and validated with dierent images sizes and performed very well by achieving high accuracy. Experimental results show that, achieved accuracy of HLReCapNet surpassed other existing network and benchmarked machine learning algorithms.

To deal with the trade-off between Self-Embedded Image Quality, Reconstruction Reference Quality and Tolerable Tampering Rate (TTR) for authentication into digital image authentication schemes, Content prioritization based selfembedding has been proposed in this paper. Image content is prioritized on a scale of 0 to 3 prior to performing the embedding. Self-embedding into each block would then be performed on the basis of its assigned priority. Higher the priority value of a block, lesser would be the embedded information in that block and thus lesser would the imposed distortion due to embedding. The prioritization of the image content also helps in generalizing the self-embedding process. Experimental results show that the proposed scheme not only allows for high quality and high protection to high priority region but also the overall quality of the self-embedded image improved as compared to the previous schemes. Also, chosen DWT-SPHIT as reference generation mechanism allows for good quality reconstruction even at low bit rates and this in turn allows more information to be embedded into the image which means lower payload, that results in successful restoration even under high tampering conditions. Furthermore, modeling of tampering problem as an erasure channel ensures high tampering tolerance.

We are witnessing a transition era of cloud security, as cloud computing paradigm is shifting its focus from provider to the consumer. Cloud service trust manipulation detection is different from traditional on-site service trust detection because cloud performs data operations at diverse geographically remote data centers; thus diminishing consumer control over the kind of service to be utilized. When the Cloud user submits a particular job to cloud, user has to rely upon the good behavior of the cloud to perform the task without violating the user trust in services utilized. However, due to lack of transparency in cloud, consumers find it hard to evaluate trust. Inspired by the recent progress of Spatial Rich Models (SRM) in image forensics domain, we propose to employ SRM and Machine learning approach to verify the trusted behavior of cloud by analyzing the rich features of the output produced by cloud service. We investigated noise distributions in data for violation detection. The approach is based on the hypothesis that every data processing task leaves certain distinct traces on the data. We identify those digital footprints to analyze whether cloud service provider has utilized the legitimate software-as-a-service for processing consumer requests. The inconsistency between authentic and obtained output acts as a proof-of-work for trust violation detection. The experimental results for the standard image dataset demonstrate that noise distributions in spatial domain can be successfully utilized to detect Cloud service trust violations.