Blind steganalysis-driven secure transmission validation using feature-based classification in JPEG images

From the digital forensics point of view, image forgery is considered as evidence that could provide a major breakthrough in the investigation process. Additionally, the development of storage device technologies has increased storage space significantly. Thus a digital investigator can be overwhelmed by the amount of data on storage devices that needs to be analysed. In this paper, we propose a model for classifying bulk JPEG images produced by the data carving process or other means into three different classes to solve the problem of identifying forgery quickly and effectively. The first class is JPEG images that contain errors or corrupted data, the second class is JPEG images that contain forged regions, and the third is JPEG images that have no signs of corruption or forgery. To test the proposed model, some experiments were conducted on our own dataset in addition to CASIA V2 image forgery dataset. The experiments covered different types of forgery technique. The results yielded around 88% accuracy rate in the classification process using five different machine learning methods on CASIA V2 dataset. It can be concluded that the proposed model can help investigators to automatically classify JPEG images, which reduce the time needed in the overall digital investigation process.

Holistic representations of natural scenes are an effective and powerful source of information for semantic classification and analysis of images. Despite the technological hardware and software advances, consumer single-sensor imaging devices technology are quite far from the ability of recognising scenes and/or to exploit the visual content during (or after) acquisition time. The frequency domain has been successfully exploited to holistically encode the content of natural scenes in order to obtain a robust representation for scene classification. The authors exploit a holistic representation of the scene in the discrete cosine transform domain fully compatible with the JPEG format. The advised representation is coupled with a logistic classifier to perform classification of the scene at superordinate level of description (e.g. natural against artificial), or to discriminate between multiple classes of scenes usually acquired by a consumer imaging device (e.g. portrait, landscape and document). The proposed method is able to work in constrained domain. Experiments confirm the effectiveness of the proposed method. The obtained results closely match state-of-the-art methods in terms of accuracy outperforming in terms of computational resources.

Since JPEG images have been widely used in our daily life, the steganalysis for JPEG images becomes very important and significant. The Article addresses steganalysis in static images based on DCT transformed region, able to recognize the most popular steganography algorithms occurring on the Internet. We propose a new steganalysis method, where statistical properties of the image are explored, regardless the embedding procedure employed. The feature set used for classification of images consists of 285 statistical features. Experimental results show that in comparing with the universal steganalysis method for JPEG stego images, our method improves detection of widely used steganographic method in detection process, which provides observable differences in investigation performance.

Adversarial attacks involve adding, small, often imperceptible, perturbations to inputs with the goal of getting a machine learning model to misclassifying them. While many different adversarial attack strategies have been proposed on image classification models, object detection pipelines have been much harder to break. In this paper, we propose a novel strategy to craft adversarial examples by solving a constrained optimization problem using an adversarial generator network. Our approach is fast and scalable, requiring only a forward pass through our trained generator network to craft an adversarial sample. Unlike in many attack strategies we show that the same trained generator is capable of attacking new images without explicitly optimizing on them. We evaluate our attack on a trained Faster R-CNN face detector on the cropped 300-W face dataset where we manage to reduce the number of detected faces to 0.5% of all originally detected faces. In a different experiment, also on 300-W, we demonstrate the robustness of our attack to a JPEG compression based defense typical JPEG compression level of 75% reduces the effectiveness of our attack from only 0.5% of detected faces to a modest 5.0%.

Deep learning models, particularly transformer architectures, have revolutionized various computer vision tasks, including image classification. However, their performance under different types and levels of noise remains a crucial area of investigation. In this study, we explore the noise sensitivity of prominent transformer models trained on the ImageNet dataset. We systematically evaluate 22 transformer variants, ranging from state-of-the-art large-scale models to compact versions tailored for mobile applications, under five common types of image distortions. Our findings reveal diverse sensitivities across different transformer architectures, with notable variations in performance observed under additive Gaussian noise, multiplicative Gaussian noise, Gaussian blur, salt-and-pepper noise, and JPEG compression. Interestingly, we observe a consistent robustness of transformer models to JPEG compression, with top-5 accuracies exhibiting higher resilience to noise compared to top-1 accuracies. Furthermore, our analysis highlights the vulnerability of mobile-oriented transformer variants to various noise types, underscoring the importance of noise robustness considerations in model design and deployment for real-world applications. These insights contribute to a deeper understanding of transformer model behavior under noisy conditions and have implications for improving the robustness and reliability of deep learning systems in practical scenarios.

Recent studies have shown that neural network (NN) based image classifiers are highly vulnerable to adversarial examples, which poses a threat to security-sensitive image recognition task. Prior work has shown that JPEG compression can combat the drop in classification accuracy on adversarial examples to some extent. But, as the compression ratio increases, traditional JPEG compression is insufficient to defend those attacks but can cause an abrupt accuracy decline to the benign images. In this paper, with the aim of fully filtering the adversarial perturbations, we firstly make modifications to traditional JPEG compression algorithm which becomes more favorable for NN. Specifically, based on an analysis of the frequency coefficient, we design a NN-favored quantization table for compression. Considering compression as a data augmentation strategy, we then combine our model-agnostic preprocess with noisy training. We fine-tune the pre-trained model by training with images encoded at different compression levels, thus generating multiple classifiers. Finally, since lower (higher) compression ratio can remove both perturbations and original features slightly (aggressively), we use these trained multiple models for model ensemble. The majority vote of the ensemble of models is adopted as final predictions. Experiments results show our method can improve defense efficiency while maintaining original accuracy.

Digital image forgery and its forensics have emerged as a significant research domain. Digital forensics is required to examine the questioned images and classify them as authentic or tampered. This paper aims at image tamper detection using a novel energy deviation measure (EDM). The EDM is a measure of deviation in pixel intensity with respect to its immediate and distant neighbourhood. It is extracted by measuring the inter pixel intensity difference across and inside the DCT block boundary of a JPEG image. Features from EDM have been used for the classification of the authentic and tampered images. Support vector machine is used for image classification. The experimental results have shown that the proposed method performs better with fewer dimensions as compared to other state of the art methods. It gives improved accuracy and area under curve while classifying images. It is robust to noise and JPEG image compression quality factor.

Illegal data access has become prevalent in wireless and general communication networks. There is a need for secure transmission of medical images since telemedicine is increasingly being used. Recently, the use of chaotic signals for secure data transmission has seen significant growth in developing chaotic encryption and decryption algorithms. Moreover, retaining the details of the medical image is particularly important for accurate diagnosis. Hence, in this paper, a secure scheme for medical image transmission is proposed. In this scheme, an enhanced chaotic key based algorithm (ECKBA) is proposed for encryption and decryption and lossless JPEG is used for compression of medical images. Further, the cryptanalysis of the proposed ECKBA is performed. The efficacy of the proposed scheme is illustrated through implementation results on a CT scanned abdomen image of a kidney patient and on MRI image of a patient with brain tumor.

Disentangling the intrinsic feature from the related feature in image classification using knowledge distillation and object replacement

Knowledge distillation aims to distill knowledge from teacher networks to train student networks. Distilling intermediate features has attracted much attention in recent years as it can be flexibly applied in various fields such as image classification, object detection and semantic segmentation. A critical obstacle of feature-based knowledge distillation is the dimension gap between the intermediate features of teacher and student, and plenty of methods have been proposed to resolve this problem. However, these works usually implement feature uniformization in an unsupervised way, lacking guidance to help the student network learn meaningful mapping functions in the uniformization process. Moreover, the dimension uniformization process of the student and teacher network is usually not equivalent as the mapping functions are different. To this end, some factors of the feature are discarded during parametric feature alignment, or some factors are blended in some non-parametric operations. In this paper, we propose a novel semantic-aware knowledge distillation scheme to solve these problems. We build a standalone feature-based classification branch to extract semantic-aware knowledge for better guiding the learning process of the student network. In addition, to avoid the inequivalence of feature uniformization between teacher and student, we design a novel parameter-free self-attention operation that can convert features of different dimensions into vectors of the same length. Experimental results show that the proposed knowledge distillation scheme outperforms existing feature-based distillation methods on the widely used CIFAR-100 and CINIC-10 datasets.

Assessment of subvisible particles in biopharmaceuticals with image feature extraction and machine learning

The JPEG compression format has been the standard for lossy image compression for over multiple decades, offering high compression rates at minor perceptual loss in image quality. For GPU-accelerated computer vision and deep learning tasks, such as the training of image classification models, efficient JPEG decoding is essential due to limitations in memory bandwidth. As many decoder implementations are CPU-based, decoded image data has to be transferred to accelerators like GPUs via interconnects such as PCI-E, implying decreased throughput rates. JPEG decoding therefore represents a considerable bottleneck in these pipelines. In contrast, efficiency could be vastly increased by utilizing a GPU-accelerated decoder. In this case, only compressed data needs to be transferred, as decoding will be handled by the accelerators. In order to design such a GPU-based decoder, the respective algorithms must be parallelized on a fine-grained level. However, parallel decoding of individual JPEG files represents a complex task. In this paper, we present an efficient method for JPEG image decompression on GPUs, which implements an important subset of the JPEG standard. The proposed algorithm evaluates codeword locations at arbitrary positions in the bitstream, thereby enabling parallel decompression of independent chunks. Our performance evaluation shows that on an A100 (V100) GPU our implementation can outperform the state-of-the-art implementations libjpegturbo (CPU) and nvJPEG (GPU) by a factor of up to 51 (34) and 8.0 (5.7). Furthermore, it achieves a speedup of up to 3.4 over nv JPEG accelerated with the dedicated hardware JPEG decoder on an A100.

Digital Whole Slide Imaging (WSI) systems allow scanning complete probes at microscopic resolutions, making image compression inevitable to reduce storage costs. While lossy image compression is readily incorporated in proprietary file formats as well as the open DICOM format for WSI, its impact on deep-learning algorithms is largely unknown. We compare the performance of several deep learning classification architectures on different datasets using a wide range and different combinations of compression ratios during training and inference. We use ImageNet pre-trained models, which is commonly applied in computational pathology. With this work, we present a quantitative assessment on the effects of repeated lossy JPEG compression for ImageNet pre-trained models. We show adverse effects for a classification task, when certain quality factors are combined during training and inference.KeywordsWhole Slide ImagingCompression artifactsPathology image classification

Stiskanje z izgubami se vse pogosteje uporablja v daljinskem zaznavanju, eprav njegov vpliv na rezultate obdelav posnetkov e ni v celoti raziskan. V lanku so predstavljeni uinki stiskanja JPEG2000 z izgubami na klasifikacijo posnetkov zelo visoke loljivosti WorldView-2. Za klasifikacijo stisnjenih posnetkov smo prvi uporabili objektni metodi k-najbliji sosed (k-nearest neighbor) in metodo podpornih vektorjev (support vector machine),

HSIs are widely employed in practical applications such as environmental monitoring, agriculture, military reconnaissance, and geographic mapping. Although Federated Learning (FL) allows decentralized model training without sharing raw data, transmitting model updates may still leak sensitive information. To address this issue, we proposed ADP-SSL (Adaptive Differentially Private Secure Socket Layer), a privacy-preserving FL framework that combines adaptive differential privacy with SSL-based communication. This design ensures secure transmission of model updates while reducing computational and communication overhead relative to traditional cryptographic techniques. Furthermore, to address the challenges posed by non-independent and identically distributed (non-IID) data across clients, we introduce FedEL (Federated Ensemble Learning). We evaluate the proposed approach on standard HSI datasets, including Indian Pines, Pavia University, and Barley. Experimental results show that ADP-SSL effectively mitigates privacy risks with minimal performance degradation, while FedEL improves classification accuracy in non-IID settings. Together, these methods offer a scalable and secure solution for real-world, distributed HSI classification.