Copy-move forgery is one of the most commonly used manipulations for tampering digital images. Keypoint-based detection methods have been reported to be very effective in revealing copy-move evidence due their robustness against various attacks, such as large-scale geometric transformations. However, these fail handle cases when forgeries only involve small or smooth regions, where number keypoints limited. To tackle this challenge, we propose a fast and algorithm through hierarchical...

In recent years, online social networks (OSNs) have become extremely popular and been one of the most common ways for storing distributing images. Naturally, such widespread availability OSN makes it a viable channel transmitting additional data along with image sharing. However, various lossy operations, e.g., resizing compression, conducted by platforms impose great challenges designing robust watermarking scheme over shared this paper, we tackle challenge propose high-capacity technique,...

Deep learning (DL) has demonstrated its powerful capabilities in the field of image inpainting. The DL-based inpainting approaches can produce visually plausible results, but often generate various unpleasant artifacts, especially boundary and highly textured regions. To tackle this challenge, work, we propose a new end-to-end, two-stage (coarse-to-fine) generative model through combining local binary pattern (LBP) network with an actual network. Specifically, first LBP using U-Net...

Predicting the motion trajectories of moving agents in complex traffic scenes, such as crossroads and roundabouts, plays an important role cooperative intelligent transportation systems. Nevertheless, accurately forecasting behavior a dynamic scenario is challenging due to interactions between agents. Graph Convolutional Neural Network has recently been employed deal with Despite promising performance resulting trajectory prediction algorithms, many existing graph-based approaches model...

Deep neural networks (DNNs) have been shown to be vulnerable against adversarial examples (AEs), which are maliciously designed cause dramatic model output errors. In this work, we reveal that normal (NEs) insensitive the fluctuations occurring at highly-curved region of decision boundary, while AEs typically over one single domain (mostly spatial domain) exhibit exorbitant sensitivity on such fluctuations. This phenomenon motivates us design another classifier (called dual classifier) with...

The pedestrian trajectory prediction task is an essential component of intelligent systems. Its applications include but are not limited to autonomous driving, robot navigation, and anomaly detection monitoring Due the diversity motion behaviors complex social interactions among pedestrians, accurately forecasting their future challenging. Existing approaches commonly adopt generative adversarial networks (GANs) or conditional variational autoencoders (CVAEs) generate diverse trajectories....

Click-through rate (CTR) prediction is a crucial task in recommender systems, which aims to model users' dynamic preferences from their historical behaviors. To achieve this goal, most of the previous models adopt sequential neural networks (e.g., GRU) encode interactions into item representations for recommendations. Though these methods can perform well on recommending highly relevant items users, we argue that such are sub-optimal long-term user experience due skewed recommendations:...

Automatic classification and segmentation of medical images play essential roles in computer-aided diagnosis. Deep convolutional neural networks (DCNNs) have shown their advantages on image segmentation. However, they not achieved the same success as done natural images. In this paper, two challenges are exploited for DCNNs images, including 1) lack feature diversity; 2) neglect small lesions. These issues heavily influence performances. To improve performance DCNN similarity-aware attention...

In most of the existing representation learning frameworks, noise contaminating data points is often assumed to be independent and identically distributed ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.i.d.</i> ), where Gaussian distribution imposed. This assumption, though greatly simplifies resulting problems, may not hold in many practical scenarios. For example, face usually attributable local variation, random occlusion,...

Recently, deep convolutional neural networks have been applied to image compressive sensing (CS) improve reconstruction quality while reducing computation cost. Existing learning-based CS methods can be divided into two classes: sampling at single scale and across multiple scales. However, these existing treat the low-frequency high-frequency components equally, which is an obstruction get a high quality. This paper proposes adaptive multi-scale network in wavelet domain called AMS-Net,...

As a prevalent form of multimodal data, video data plays crucial role in numerous applications, offering various benefits. Meanwhile, integrity and source issues also pose security risks. Video is multimodal, containing container describing coding packaging, along with stream featuring visual audio information. Many works on analysis focus containers, they overlook the fact that malicious user can readily manipulate these traces within containers by reconstructing them without transcoding....

Image inpainting based on generative adversarial networks (GANs) has achieved great success in producing visually plausible images and plays an important role many real tasks. However, the techniques of image might also be maliciously used, e.g., altering or removing interesting objects to report fake news. Despite promising performance recently developed detection algorithms, they are built convolutional neural (CNNs) with limited receptive fields. Consequently, fail fully capture disparity...

Scale invariant feature transform (SIFT), as one of the most popular local extraction algorithms, has been widely employed in many computer vision and multimedia security applications. Although SIFT extensively investigated from various perspectives, its against malicious attacks rarely discussed. In this paper, we show that keypoints can be effectively removed with minimized distortion on processed image. The keypoint removal is formulated a constrained optimization problem, where...

Most of the existing subspace clustering (SC) frameworks assume that noise contaminating data is generated by an independent and identically distributed (i.i.d.) source, where Gaussianity often imposed. Though these assumptions greatly simplify underlying problems, they do not hold in many real-world applications. For instance, face clustering, usually caused random occlusions, local variations unconstrained illuminations, which essentially structural hence satisfies neither i.i.d. property...

Screen-shooting resilient (SSR) watermark is a special kind of robust watermarking. One can extract the message even embedded image communicates via physical screen to camera channel. The keypoint-based SSR watermarking one promising solution realize such screen-to-camera communication. enhanced keypoints were used locate embedding region and then perform embedding. However, treats critical two steps, keypoint enhancement embedding, independently, neglecting their inter-play. This work...

Cloud computing offers advantages in handling the exponential growth of images but also entails privacy concerns on outsourced private images. Reversible data hiding (RDH) over encrypted has emerged as an effective technique for securely storing and managing confidential cloud. Most existing schemes only work uncompressed However, almost all are transmitted stored compressed formats such JPEG. Recently, some RDH JPEG bitstreams have been developed, these works disadvantages a small embedding...