Deepfake technology poses a serious threat to society by synthesizing a victims facial features and attributes to carry out deception. Traditional active defense methods against deepfake attacks are typically designed for specific models, and protected images often lose their anti-forgery capability after compression or reconstruction, severely limiting their practical applicability. This paper proposes a Meta-Learning-based active defense Scheme against deep facial forgery attacks (MLPDS), which effectively safeguards facial images against diverse deepfake attacks in real-world scenarios. Our approach adopts a general paradigminjecting noise into the original image to construct a cross-model defense algorithm against deepfake attacks. Specifically, by leveraging a meta-learning strategy, we integrate perturbations generated by multiple deepfake models, enabling robust protection against a variety of forgery models. Furthermore, to maintain the high fidelity of the images, we propose a symmetric gradient quantization strategy based on the arctan function to minimize the perceptual discrepancy between the perturbed and original images. Finally, an end-to-end optimization network is employed to generate universal perturbations tailored to specific images, supported by a pixel-level error metric that constrains deviations from the original content. Since no retraining is required to protect newly encountered images, this approach significantly improves the efficiency and practicality of real-time anti-deepfake defense. Experiments show that the proposed MLPDS algorithm can effectively resist attacks from multiple forgery models, outperforming state-of-the-art defense methods and significantly reducing image distortion with an average PSNR gain of approximately 7 dB, which fully meets the practical desire for efficient and reliable deepfake defense.

Existing image steganography schemes always introduce obvious modification traces to the cover image, resulting in the risk of secret information leakage. To address this issue, an end-to-end framework for joint makeup style transfer and image steganography is proposed in this paper to achieve imperceptible higher-capacity data hiding. In the scheme, a Parsing-guided Semantic Feature Alignment (PSFA) module is designed to transfer the style of a makeup image to an object non-makeup image, thereby generating a content-style integrated feature matrix. Meanwhile, a Multi-Scale Feature Fusion and Data Embedding (MFFDE) module was devised to encode the secret image into its latent features and fuse them with the generated content-style integrated feature matrix, as well as the non-makeup image features across multiple scales, to achieve the makeup-stego image. As a result, the style of the makeup image is well transformed and the secret image is imperceptibly embedded simultaneously without directly modifying the pixels of the original non-makeup image. Additionally, a Residual-aware Information Compensation Network (RICN) is developed to compensate the loss of the secret image arising from the multilevel data embedding, thereby further enhancing the quality of the reconstructed secret image. Experimental results show that the proposed scheme achieves superior steganalysis resistance capability and visual quality in both makeup-stego images and recovered secret images, compared with other state-of-the-art schemes.

The images transmitted by IoT devices, particularly those used for surveillance or sensor data, are vulnerable to malicious screenshots and unauthorized access, leading to potential privacy breaches. To address this, we propose a multi-stage Singular Value Decomposition (SVD)-based robust data-hiding scheme for JPEG images aimed at mitigating screenshot attacks. The method exploits the decorrelation properties of the Discrete Cosine Transform (DCT) to preprocess the carrier image, facilitating the selection of specific frequency coefficients. These coefficients undergo a dual-stage SVD transformation, where dimensionality reduction reduces the impact of noise from non-critical image regions. Additionally, we optimize Grad-CAM heatmap generation to better align with human visual perception, enabling the identification of stable and reliable feature regions for embedding secret information. This approach ensures that the visual integrity of the carrier image is maintained while preserving the legibility of the embedded information, even under attack.Our method enhances both the visual fidelity of the carrier image and the robustness of the embedded information. Experimental results demonstrate that the proposed scheme outperforms existing methods, achieving at least a 5% improvement in confidential information extraction accuracy and a data extraction rate exceeding 95% across screenshot angles ranging from -40∘ to 40∘. Extensive evaluations confirm the superior performance, efficiency, and security of our method in mitigating screenshot attacks, showcasing its broad applicability to various image formats and resilience to distortions.

Perceptual image hashing has emerged as a crucial forensic tool within the Internet of Things (IoT) ecosystem. Traditional perceptual hashing algorithms predominantly rely on global image features to generate hash codes, which limit their ability to represent key features of images effectively. This paper introduces a Perceptual Region Recognition Network (PRRN) to accurately identify key feature regions in images based on their texture distribution characteristics, thereby generating image perceptual hashing codes that reflect the key content of the images. At the same time, a perceptual hashing feature extraction module, which integrates a Residual Network (ResNet) and a Weighted Feature Fusion Network (WFFN), is built to extract deep semantic features of the object image. Where, ResNet is leveraged to extract high-level semantic features, while WFFN ensures the preservation of low-level local features. Furthermore, skip connections are employed to achieve content enhancements for intricate details of critical image regions. Additionally, the Mean Squared Error (MSE) loss is incorporated to enhance the accuracy of key region localization, further improving the sensitivity of image perceptual hash codes and accelerating the network’s convergence speed. Extensive experimental evaluations demonstrate that the proposed PRRN-based perceptual image hashing scheme significantly outperforms other state-of-the-art methods in terms of image feature representation capability. Specifically, it achieves an average improvement of over 1.2 in attack-resistant capability for images compared with other counterparts, making it a promising candidate for practical applications in the IoT environment.

The proliferation of DeepFakes has heightened the necessity to distinguish between authentic and counterfeit faces. While numerous methods concentrate on detecting DeepFakes, only a few address safeguarding genuine faces from manipulation. This letter proposes a novel active forensics system for DeepFake forensics utilizing separable perceptual hash enhancement algorithm. A separable perceptual hash code specifically designed for face deep forgery is introduced, achieving robustness while maintaining sensitivity and imperceptibility when embedded within the original image. Additionally, a multi-scale perceptual smoothing loss function is employed to optimize perceptual similarity, structural smoothness, and embedding stability. As a result, this system ensures the consistence of confidential information both before and after manipulation, thereby enhancing the capability of face source detection and DeepFake identification. Experimental results demonstrate that the proposed scheme can effectively counter traditional deep forgery techniques while exhibiting significant potential in preserving personal privacy.

The photo response non-uniformity (PRNU) is used to connect an image to its source sensor. In this paper, researchers propose a PRNU anonymity method based on image segmentation to cut the relationship between the image and its source camera. According to the distribution rule of PRNU in the high and low frequency band of the image, the high and low frequency information of the part is also processed differently, which ensures the quality of the output image to a large extent. Experiments on the datasets show that the proposed method can preserve the biometric characteristics of the device while maintaining the anonymity of the device. Comparing with prior art, peak signal to noise ratio (PSNR) and cosine similarity are improved by 1.9 dB and 0.02 points, respectively.

Channel estimation is important for orthogonal frequency division multiplexing (OFDM) in current wireless communication systems. Prevalent channel estimation algorithms, however, cannot be widely deployed due to some practical reasons, such as poor robustness and high computational complexity. To solve the problems for OFDM systems, we propose a new channel estimation scheme with a fine-designed deep learning model, called RRDBNet. RRDBNet can be trained easily while maintaining the advantages of residual learning and increasing the structure capacity, by combining the multi-level residual network and dense links. Our simulation results show that RRDBNet outperforms the traditional least-square algorithm and existing DL-based super-resolution schemes, which ranges from 0.5 to 1dB at low SNR and from 2 to 3dB at high SNR. Besides, in terms of the number of pilots, RRDBNet is also superior to existing schemes and approaches LMMSE.

At present, there is no unified network resource scheduling and business optimization system in cloud-edge collaboration services with supercomputing as the core, especially network transmission and data management. In this poster, we focus on how to optimize the quality of user experience while meeting huge computing power requirements, and design a novel network resource scheduling and traffic optimization platform centered on supercomputing system, which is called PANGU. PANGU is a multi-dimensional network resource scheduling solution for supercomputing that can guarantee collaborative network services. PANGU can make up for the lack of computing power of edge computing while solving the adaptability of supercomputing nodes to pervasive cloud edge computing. Through effective network resource management, PANGU can provide higher productivity.

The cloud-edge integration service model combines the advantages of computing capabilities both from cloud and edge. Therefore, the data centers with cloud-edge integrated are an irreversible trend for the evolution of future data center. Software-Defined Network (SDN), emerging as a novel network model, separates content forwarding and control, and that makes resource management across data center network more efficient. This article focuses on the network data transmission and management for future data centers. First, it reviews measurement, analysis, and monitoring methods meant for new features of global SDN network. Then it focuses on unified management of SDN resources like traffic scheduling theory for cross-domain data centers based on cloud. Specifically, we proposed a novel fault response mechanism across the network with a more precise location and less response time. With dynamic changes of cloud computing and edge computing services combined, global QoS control and QoE optimization methods are proposed correspondingly. Finally, a set of SDN control platforms supporting the functions mentioned above are formulated. We hope that our work will shed some new light and provide new theoretical support for cloud-edge-combined cross-domain data center network architecture.

The transmission of medical image data in an open network environment is subject to privacy issues including patient privacy and data leakage. In the past, image encryption and information-hiding technology have been used to solve such security problems. But these methodologies, in general, suffered from difficulties in retrieving original images. We present in this paper an algorithm to protect key regions in medical images. First, coefficient of variation is used to locate the key regions, a.k.a. the lesion areas, of an image; other areas are then processed in blocks and analyzed for texture complexity. Next, our reversible data-hiding algorithm is used to embed the contents from the lesion areas into a high-texture area, and the Arnold transformation is performed to protect the original lesion information. In addition to this, we use the ciphertext of the basic information about the image and the decryption parameter to generate the Quick Response (QR) Code to replace the original key regions. Consequently, only authorized customers can obtain the encryption key to extract information from encrypted images. Experimental results show that our algorithm can not only restore the original image without information loss, but also safely transfer the medical image copyright and patient-sensitive information.

Obstructive sleep apnea syndrome (OSAS) is a breathing disorder presenting during sleep. Although polysomnography (PSG) is the gold standard to diagnose OSAS, it is an expensive method that is quite complicated to use. Worse, it takes a long time between testing and getting a diagnosis from PSG. Thus, we have designed an algorithm aimed at diagnosing OSAS in a more efficient manner. First, blood oxygen saturation (SpO2) data are processed to obtain statistical features, which are then trained to establish a classification model based on a support vector machine (SVM) strategy; the resulting SVM model performs the diagnosis of OSAS. Furthermore, in order to allow remote diagnosis, we combine our algorithm with a monitoring system. To achieve this, physiological data are collected from a smart phone and then uploaded to the SVM model in the cloud. Once processed, a diagnosis report is returned to the smart phone. A preliminary evaluation of our algorithm based on real-world data is extremely promising as we find its accuracy, sensitivity, and specificity to be 90.2%, 87.6%, and 94.1%, respectively.