Neural Networks最新文献

Schizophrenia (SZ) is characterized by cognitive impairments and widespread structural brain alterations. The potential adaptability of convolutional neural networks (CNN) to identify the complex and extensive brain alterations associated with SZ relies on its automatic feature learning capability. Structural magnetic resonance imaging (sMRI) is a non-invasive technique for investigating disruptions related to white matter (WM), grey matter (GM), and cerebrospinal fluid (CSF) of brain regions. We proposed an intrinsic CNN ensemble of kernel ridge regression-based random vector functional link (KRR-RVFL) architecture to explore the WM disruptions for SZ. In this approach, we have integrated an eight-layer CNN into five different KRR-RVFL classifiers for feature extraction and classification. The classifiers' outputs are averaged and fed to the final KRR-RVFL classifier for final classification. The KRR-RVFL classifier enhances stability and robustness by addressing non-linearity limitations in the standard RVFL network. The proposed CNN ensemble KRR-RVFL outperforms other classifiers with 97.33 % accuracy for the WM region, showing significant disruptions compared to GM and CSF. Furthermore, we calculated the correlation coefficient between tissue volumes and the scale of symptoms for GM and WM. According to the results, tissue volume for WM is reduced more than GM for SZ. The proposed model assists clinicians in exploring the role of WM disruptions for accurate diagnosis of SZ.

Text-to-Image (T2I) diffusion models have gained significant traction due to their remarkable image generation capabilities, raising growing concerns over the security risks associated with their use. Prior studies have shown that malicious users can subtly modify prompts to produce visually misleading or Not-Safe-For-Work (NSFW) content, even bypassing existing safety filters. Existing adversarial attacks are often optimized for specific prompts, limiting their generalizability, and their text-space perturbations are easily detectable by current defenses. To address these limitations, we propose a universal adversarial attack framework called dormant key. It appends a transferable suffix that can be appended as a "plug-in" to any text input to guide the generated image toward a specific target. To ensure robustness across diverse prompts, we introduce a novel hierarchical gradient aggregation strategy that stabilizes optimization over prompt batches. This enables efficient learning of universal perturbations in the text space, improving both attack transferability and imperceptibility. Experimental results show that our method effectively balances attack performance and stealth. In NSFW generation tasks, it bypasses major safety mechanisms, including keyword filtering, semantic analysis, and text classifiers, and achieves over 18 % improvement in success rate over baselines.

The rapid development of generative models has improved image quality and made image synthesis widely accessible, raising concerns about content credibility. To address this issue, we propose a method called Universal Reconstruction Residual Analysis (UR²EA) for detecting synthetic images. Our study reveals that, when GAN- and diffusion-generated images are reconstructed by pre-trained diffusion models, they exhibit significant differences in reconstruction error compared to real images: GAN-generated images show lower reconstruction quality than real images, whereas diffusion-generated images are more accurately reconstructed. We leverage these residual maps as a universal prior to training a model for detecting synthetic images. In addition, we introduce a Multi-scale Channel and Window Attention (MCWA) module to extract fine-grained features from residual maps across multiple scales, capturing both local and global details. To facilitate the exploration of diverse detection methods, we constructed a new UniversalForensics dataset, which includes various representations of synthetic images generated by 30 different models. Compared to the best-performing baselines, our method improves average accuracy by 3.3 % and precision by 1.6 %, achieving state-of-the-art results.