Median filtering is used as an anti-forensic technique to erase processing history of some image manipulations such JPEG, resampling, etc. Thus, various detectors have been proposed detect median filtered images. To counter these techniques, several methods devised well. However, restoring the a typical ill-posed problem, and thus it still difficult reconstruct visually close original image. Also, further hard make restored statistical characteristic raw for purpose. solve this we present...

As technological developments have enabled high-quality fingerprint scanning, sweat pores, one of the Level 3 features fingerprints, been successfully used in automatic recognition systems (AFRS). Since pore extraction process is a critical step for AFRS, high accuracy required. However, it difficult to extract correctly because shape depends on person, region, and type. To solve problem, we presented method using deep convolutional neural networks intensity refinement. The are detect pores...

In this paper, we propose a composite manipulation detection method based on convolutional neural networks (CNNs). To our best knowledge, is the first work applying deep learning for forgery detection. The proposed technique defines three types of attacks that occurred frequently during image forging and detects when they are concurrently applied to images. do this, learn statistical change due through CNN architecture classify manipulated image. effective since it learns integrated extracts...

The current research direction in generative models, such as the recently developed GPT4, aims to find relevant knowledge information for multimodal and multilingual inputs provide answers. Under these circumstances, demand evaluation of visual question answering (VQA) tasks, a representative task systems, has increased. Accordingly, we propose bilingual outside-knowledge VQA (BOK-VQA) dataset this study that can be extended multilingualism. proposed data include 17K images, question-answer...

Various watermarking schemes achieved the robustness against usual operations such as simplification, remeshing and noise addition. However, techniques were not robust cropping, nevertheless cropping attack is commonly performed by general editing. In this paper, we propose a 3D mesh method cropping. We achieve blind scheme involving consistent segmentation improves The experimental results show that proposed achieves higher performance than previous methods.

Abstract Image steganalysis is the task of detecting a secret message hidden in an image. Deep using end-to-end deep learning has been successful recent years, but previous studies focused on improving detection performance rather than designing lightweight model for practical applications. This caused to be heavy and computationally costly, making infeasible deploy real-world To address this issue, we study effective design strategy image steganalysis. Considering domain-specific...

Steganalysis refers to the study of identifying hidden messages in images inserted by steganography. Although detection performance is greatly improved when adopting convolutional neural networks (CNNs), they require sophisticated tricks, such as preprocessing for suppression image content, using absolute and truncated activation functions, utilizing domain knowledge. These tricks help train stably mitigate convergence problem early stages training, but also restrict flexibility CNNs, which...

In digital image forensics, previous methods for hue forgery detection cannot be used after common processing such as resizing and JPEG compression. this paper, we suggest a robust forensics scheme estimating modification of images. To achieve goal, use sensor pattern noise from each color channel un-tampered images the ground truth. Since know unique characteristics channel, can estimate by testing suspicious all changes. The results confirms that proposed method distinguishes estimates...

Several depth image-based rendering (DIBR) watermarking methods have been proposed, but they various drawbacks, such as non-blindness, low imperceptibility and vulnerability to signal or geometric distortion. This paper proposes a template-based DIBR method that overcomes the drawbacks of previous methods. The proposed exploits two properties resist attacks: pixel is only moved horizontally by DIBR, smaller block not distorted DIBR. one-dimensional (1D) discrete cosine transform (DCT)...

Abstract Visualization of pathlines is common and highly relevant for the analysis unsteady flow. However, can intersect, leading to visual clutter perceptual issues. This makes it intrinsically difficult provide expressive visualizations entire domain by an arrangement multiple pathlines, in contrast well‐established streamline placement techniques. We present approach reduce these problems. It inspired glyph‐based visualization small multiples: we partition into cells, each corresponding a...

A digital certificate under Public Key Infrastructure has a defect of Man-in-the-Middle Attack that performs hash collision attacks. In this paper, we propose robust biometric-PKI authentication system against Attack. The consists current PKI and biometric authentication, which employs data public key from certificate. the proposed system, an au- thentication process it extracts consistent features fingerprint images, encrypts features, matches with prepared templates. simulation results...

Since convolutional neural networks have shown remarkable performance on various computer vision tasks, many network architectures for image steganalysis been introduced. Many of them use fixed preprocessing filters stable learning, which a disadvantage limited the information input image. The recently introduced end-to-end learning method uses structure that limits number channels feature maps close to and stacks residual blocks. This has limitations in generating levels resolutions can be...

Several depth image based rendering (DIBR) watermarking methods have been proposed, but they various drawbacks, such as non-blindness, low imperceptibility, and vulnerability to signal or geometric distortion. This paper proposes a template DIBR method that overcomes the drawbacks of previous methods. The proposed exploits two properties resist attacks: pixel is only moved horizontally by DIBR, smaller block not distorted DIBR. one dimensional (1D) discrete cosine transform (DCT) curvelet...

The current research direction in generative models, such as the recently developed GPT4, aims to find relevant knowledge information for multimodal and multilingual inputs provide answers. Under these circumstances, demand evaluation of visual question answering (VQA) tasks, a representative task systems, has increased. Accordingly, we propose bilingual outside-knowledge VQA (BOK-VQA) dataset this study that can be extended multilingualism. proposed data include 17K images, question-answer...

스테가노그래피는 비밀 메시지를 은닉하는 기술로써 기밀사항을 외부로 유출시키거나 멀웨어를 용도로 사용될 수 있기 때문에 이에 대한 탐지 기술이 중요해지고 있다. 최근 딥러닝 기반의 스테가노그래피 등장하였고 기존 특징점 기반 기술에 비해 높은 성능을 보이고 그러나 모델은 복잡성과 연산 비용이 발생하기 실제 환경에서 사용하기에는 제약이 많다. 본 논문에서는 이러한 문제를 극복하기 위해 픽셀 히스토그램을 활용한 효율적인 기법을 제안한다. 제안하는 기법은 적은 연산량과 정확도를 보이기 활용도가 높을 것으로 기대한다.

We propose the VLR-Bench, a visual question answering (VQA) benchmark for evaluating vision language models (VLMs) based on retrieval augmented generation (RAG). Unlike existing evaluation datasets external knowledge-based VQA, proposed VLR-Bench includes five input passages. This allows testing of ability to determine which passage is useful given query, capability lacking in previous research. In this context, we constructed dataset 32,000 automatically generated instruction-following...