Hateful meme detection aims to prevent the proliferation of hateful memes on various social media platforms. Considering its impact on social environments, this paper introduces a previously ignored but significant threat to hateful meme detection: backdoor attacks. By injecting specific triggers into meme samples, backdoor attackers can manipulate the detector to output their desired outcomes. To explore this, we propose the Meme Trojan framework to initiate backdoor attacks on hateful meme detection. Meme Trojan involves creating a novel Cross-Modal Trigger (CMT) and a learnable trigger augmentor to enhance the trigger pattern according to each input sample. Due to the cross-modal property, the proposed CMT can effectively initiate backdoor attacks on hateful meme detectors under an automatic application scenario. Additionally, the injection position and size of our triggers are adaptive to the texts contained in the meme, which ensures that the trigger is seamlessly integrated with the meme content. Our approach outperforms the state-of-the-art backdoor attack methods, showing significant improvements in effectiveness and stealthiness. We believe that this paper will draw more attention to the potential threat posed by backdoor attacks on hateful meme detection.

With more event datasets being released online, safeguarding the event dataset against unauthorized usage has become a serious concern for data owners. Unlearnable Examples are proposed to prevent the unauthorized exploitation of image datasets. However, it's unclear how to create unlearnable asynchronous event streams to prevent event misuse. In this work, we propose the first unlearnable event stream generation method to prevent unauthorized training from event datasets. A new form of asynchronous event error-minimizing noise is proposed to perturb event streams, tricking the unauthorized model into learning embedded noise instead of realistic features. To be compatible with the sparse event, a projection strategy is presented to sparsify the noise to render our unlearnable event streams (UEvs). Extensive experiments demonstrate that our method effectively protects event data from unauthorized exploitation, while preserving their utility for legitimate use. We hope our UEvs contribute to the advancement of secure and trustworthy event dataset sharing. Code is available at: https://github.com/rfww/uevs.

The growing popularity of 3D Gaussian Splatting (3DGS) has intensified the need for effective copyright protection. Current 3DGS watermarking methods rely on computationally expensive fine-tuning procedures for each predefined message. We propose the first generalizable watermarking framework that enables efficient protection of Splatter Image-based 3DGS models through a single forward pass. We introduce GaussianBridge that transforms unstructured 3D Gaussians into Splatter Image format, enabling direct neural processing for arbitrary message embedding. To ensure imperceptibility, we design a Gaussian-Uncertainty-Perceptual heatmap prediction strategy for preserving visual quality. For robust message recovery, we develop a dense segmentation-based extraction mechanism that maintains reliable extraction even when watermarked objects occupy minimal regions in rendered views. Project page: https://kevinhuangxf.github.io/marksplatter.

Recently, with the popularity of social media platforms, the increase in suicide-related content has triggered great attention on automated suicide risk detection systems. This research proposes a natural language processing model based on RoBERTa, which aims to identify and classify suicide risks in social media posts. Our team propose a "Multisource data filtering and fusion optimization framework"(MDF-FOF), which effectively solves the problem of data imbalance by combining a large language model (LLM) with manual annotation to generate a balanced training sample set. In terms of model construction, we proposed a framework MDF-FOF that utilizes pre-trained RoBERTa for feature extraction and combines it with a self-built EmoBERT for soft label classification. Experimental results show that the recall rate of the proposed framework on the validation set is stable above 0.75, which is better than the existing methods and baseline models. It shows good robustness and excellent performance in tests on datasets from different sources. This study provides a reliable basic framework for multi-feature extraction of user posts in the future.

Backdoor attack aims to deceive a victim model when facing backdoor instances while maintaining its performance on benign data. Current methods use manual patterns or special perturbations as triggers, while they often overlook the robustness against data corruption, making backdoor attacks easy to defend in practice. To address this issue, we propose a novel backdoor attack method named Spy-Watermark, which remains effective when facing data collapse and backdoor defense. Therein, we introduce a learnable watermark embedded in the latent domain of images, serving as the trigger. Then, we search for a watermark that can withstand collapse during image decoding, cooperating with several anti-collapse operations to further enhance the resilience of our trigger against data corruption. Extensive experiments are conducted on CIFAR10, GTSRB, and ImageNet datasets, demonstrating that Spy-Watermark overtakes ten state-of-the-art methods in terms of robustness and stealthiness.

Face recognition is often used for biometric validation, which has become a significant technique in our society. Due to its sensitive applications, security vulnerabilities posed by backdoor attacks have attracted considerable focus. Current backdoor attack methods use digital perturbations or physical objects as triggers, while these additional requirements make existing backdoor attacks less viable in real-world applications. To address this issue, we propose a novel backdoor attack method named Scene Backdoor (Scenedoor), which injects a 3D scene as the trigger that effectively simplifies the backdoor activation. Any person who appears in this scene will be attacked as the attacker-desired identity. Specifically, we reconstruct a 3D scene from several 2D images and then blend the facial part extracted from the input sample with the reconstructed scene to generate the poisoned image. Extensive experiments are conducted on CelenDF (v2), CelebA-HQ, and PinsFace datasets, demonstrating that Scenedoor overtakes five state-of-the-art methods in terms of effectiveness, stealthiness, and robustness.

Change detection (CD) is to decouple object changes (i.e., object missing or appearing) from background changes (i.e., environment variations) like light and season variations in two images captured in the same scene over a long time span, presenting critical applications in disaster management, urban development, etc. In particular, the endless patterns of background changes require detectors to have a high generalization against unseen environment variations, making this task significantly challenging. Recent deep learning-based methods develop novel network architectures or optimization strategies with paired-training examples, which do not handle the generalization issue explicitly and require huge manual pixel-level annotation efforts. In this work, for the first attempt in the CD community, we study the generalization issue of CD from the perspective of data augmentation and develop a novel weakly supervised training algorithm that only needs image-level labels. Different from general augmentation techniques for classification, we propose the background-mixed augmentation that is specifically designed for change detection by augmenting examples under the guidance of a set of background changing images and letting deep CD models see diverse environment variations. Moreover, we propose the augmented & real data consistency loss that encourages the generalization increase significantly. Our method as a general framework can enhance a wide range of existing deep learning-based detectors. We conduct extensive experiments in two public datasets and enhance four state-of-the-art methods, demonstrating the advantages of our method. We release the code at https://github.com/tsingqguo/bgmix.

Change detection (CD) aims at finding change objects from bi-temporal images, which has wide applications in different vision tasks. Previous CD methods focus more on fusing inter-layer multiscale features while ignoring the intra-layer multiscale characteristics, which hurts the integrity of change objects with different sizes. In this paper, we propose to mix intra- and inter-layer multiscale features to generate more complete change regions. To realize intra-layer multi-scale, we propose inception difference module (IDM), which employs convolutional filters with different sizes, absolute differences, and residual connections to capture intra-layer multiscale characteristics. To capture inter-layer multiscale, we propose a residual network refinement module (RNR) to fuse the features from the highest layer to the lowest layer and generate finely detailed change predictions. Our method can capture complete changes of different sizes by considering the multiscale characteristics of intra- and inter-layer simultaneously. Experiments on two benchmark datasets reveal that our method outperforms six state-of-the-art change detectors.

Abstract Change detection (CD) is a fundamental problem that aims at detecting changed objects from two observations. Previous CNN‐based CD methods detect changes through multi‐scale deep convolutional features extracted from two images. However, we find that change always occurs in the ‘Query’ image for fixed cameras. This condition means that changes can be detected in advance from a single image with a coarse change. In this paper, we propose an efficient CD method to detect precise changes from the change image. First, a change image selector is designed to identify the image containing changes. Second, a coarse change prior map generator is proposed to generate coarse change prior to indicate the position of changes. Then, we introduce a simple multi‐scale CD module to refine the coarse change detection. As only one image is used in the multi‐scale CD module, our method is more efficient in training and testing than other compared methods. Numerous experiments have been conducted to analyse the effectiveness of the proposed method. Experimental results show that the proposed method achieves superior detection performance and higher speed than other compared CD methods.

Discovering regularities in Traditional Chinese Medicine (TCM) formula has been a hot topic in assisting TCM clinical treatment and poly-pharmacology research. Several machine learning methods, like topic model, auto-encoder, and GNNs, have been proposed for discovering regularities in TCM. However, they are often limited by specific data challenges (e.g., complex relations with rich TCM knowledge, sparsity and ambiguity, expensive data labeling, etc.) in TCM formulae. Addressing these challenges, we first establish a TCM Attributed Heterogeneous Information Network (TAHIN) for modeling massive formulae, which can assemble various types of additional information and capture their relations. Based on the TAHIN, we further propose a novel hybrid-scales graph contrastive learning framework to learn high-quality node representations in a whole unsupervised manner which can be helpful for various tasks of discovering regularities such as herb classification and herb similarity search, etc. Extensive experiments demonstrate the effectiveness and interpretability of our method. Our source code and datasets are available at https://github.com/Yonggie/ HsCTRD.

Abstract Change detection (CD) is a fundamental yet challenging problem, which aims at detecting changed object in two observations. Recent CD methods are designed based on the off‐the‐shelf semantic segmentation network architectures, which is not optimal for extracting and using change‐related features. In this paper, a novel CD network architecture is proposed, including change‐related feature extraction, cross feature enhancement, and multi‐level supervision. Absolute difference of the features of different convolutional layers is first computed from a Unet‐like network for two observations. The features are partitioned into high‐ and low‐level features according to their functionalities. Then the high‐ and low‐level features are recurrently refined by cross feature enhancement to increase the representational ability of the features. The network learns change‐related features with multi‐level supervisions. The final CD result can be obtained by fusing multiple predictions. Experimental results on three CD benchmark datasets indicate the superiority of the authors' method when compared with six state‐of‐the‐art deep learning‐based CD methods.