With the rapid development of internet applications, privacy protection secret messages in covert communication has become increasingly important. To address issue attacks received communication, such as eavesdropping attacks, steganalysis and tempering we propose a novel scheme that combines coverless steganography image transformation for messages. Instead using an carrier to embed messages, our approach hides by exploiting generative network take it latent input synthesize stego image,...

Copy-move forgery is a manipulation of copying and pasting specific patches from to an image, with potentially illegal or unethical uses. Recent advances in the forensic methods for copy-move have shown increasing success detection accuracy robustness. However, images high self-similarity strong signal corruption, existing algorithms often exhibit inefficient processes unreliable results. This mainly due inherent semantic gap between low-level visual representation high-level concept. In...

The task of text-to-image generation has achieved tremendous success in practice, with emerging concept models capable producing highly personalized and customized content. Fervor for is increasing rapidly among users, platforms sharing have sprung up. owners may upload malicious concepts disguise them non-malicious text descriptions example images to deceive users into downloading generating platform needs a quick method determine whether prevent the spread concepts. However, simply relying...

Image forensics is a rising topic as the trustworthy multimedia content critical for modern society. Like other vision-related applications, forensic analysis relies heavily on proper image representation. Despite importance, current theoretical understanding such representation remains limited, with varying degrees of neglect its key role. For this gap, we attempt to investigate forensic-oriented distinct problem, from perspectives theory, implementation, and application. Our work starts...

Photo-response non-uniformity (PRNU), as a class of device fingerprint, plays key role in the forgery detection/localization for visual media. The state-of-the-art PRNU-based forensics methods generally rely on multi-scale trace analysis and result fusion, with Markov random field model. However, such hand-crafted strategies are difficult to provide satisfactory decision, exhibiting high false-positive rate. Motivated by this, we propose an end-to-end decision fusion strategy, where mapping...

The advent of artificial intelligence-generated content (AIGC) represents a pivotal moment in the evolution information technology. With AIGC, it can be effortless to generate high-quality data that is challenging for public distinguish. Nevertheless, proliferation generative across cyberspace brings security and privacy issues, including leakages individuals media forgery fraudulent purposes. Consequently, both academia industry begin emphasize trustworthiness data, successively providing...

Despite the various privacy protection methods that are available through medical services platforms, it is still challenging for patients to achieve a desirable level of during image sharing. Therefore, this paper proposes mechanism, called PPM-SEM, secure sharing electronic patient records (EPRs) and images in telemedicine; includes two stages: preparation reconstruction. In first stage, dual watermark (i.e., an watermark) generated by combining patient's EPRs with image, which can be...

Developing robust and interpretable vision systems is a crucial step towards trustworthy artificial intelligence. In this regard, promising paradigm considers embedding task-required invariant structures, e.g., geometric invariance, in the fundamental image representation. However, such representations typically exhibit limited discriminability, limiting their applications larger-scale tasks. For open problem, we conduct systematic investigation of hierarchical exploring topic from...

A long-standing topic in artificial intelligence is the effective recognition of patterns from noisy images. In this regard, recent data-driven paradigm considers 1) improving representation robustness by adding samples training phase (i.e., data augmentation) or 2) pre-processing image learning to solve inverse problem denoising). However, such methods generally exhibit inefficient process and unstable result, limiting their practical applications. paper, we explore a non-learning that aims...

Inpainting the given region of an image is a typical requirement in computer vision. Conventional inpainting, through exemplar-based or diffusion-based strategies, can create realistic inpainted images at very low cost. Also, such easy-to-use manipulation poses new security threats. Therefore, detection inpainting has attracted considerable attention from researchers. However, existing methods are typically not suitable for general various algorithms. Motivated by this, this work, efficient...

Image recoloring is an emerging editing technique that can change the color style of image by modifying pixel values without altering original content. With rapid proliferation social network and techniques, recolored images (RIs) have raised new security issues in society. Existing detection methods good performance detecting RIs for certain categories techniques. However, on handcrafted scenario still poor due to influence human prior knowledge. To deal with this problem, we explore a...

Restoring missing areas without leaving visible traces has become a trivial task with Photoshop inpainting tools. However, such tools have potentially illegal or unethical uses, as removing specific objects in images to deceive the public. Despite emergence of many forensics methods image inpainting, their detection ability is still insufficient when attending professional inpainting. Motivated by this, we propose novel method termed primary-secondary network (PS-Net) localize inpainted...

Integrating invariance into data representations is a principled design in intelligent systems and web applications. Representations play fundamental role, where applications are both built on meaningful of digital inputs (rather than the raw data). In fact, proper design/learning such relies priors w.r.t. task interest. Here, concept symmetry from Erlangen Program may be most fruitful prior -- informally, system transformation that leaves certain property invariant. Symmetry ubiquitous,...

In the clinical analysis and processing for EEG, because of difference ages pathology, it is possible to appear abnormal waves related with pathology. Also, restrain normal rhythms could be abnormal. But at present doctors estimate if certain rhythm restrained only by using method eyeballing or some simply methods in EEG detection, which will inevitably lead errors not observable. By "the virtual record instrument" introduced this paper, all kinds characteristic waveforms (e.g. epileptic...

The vulnerability of deep neural networks to adversarial perturbations has been widely perceived in the computer vision community. From a security perspective, it poses critical risk for modern systems, e.g., popular Deep Learning as Service (DLaaS) frameworks. For protecting models while not modifying them, current algorithms typically detect patterns through discriminative decomposition natural and data. However, these decompositions are either biased towards frequency resolution or...

The effective recognition of patterns from blurred images presents a fundamental difficulty for many practical vision tasks. In the era deep learning, main ideas to cope with this are data augmentation and deblurring. However, both facing issues such as inefficiency, instability, lack explainability. paper, we explore simple but way define invariants images, without Here, designed Fractional Moments under Projection operators (FMP), where blur invariance rotation guaranteed by general...