Acquiring pairwise noisy-clean training data is challenging. Consequently, some self-supervised denoising methods utilize noisy image pairs as both input and target for network training. However, a major issue with these methods is the gap between the clean images of the input and target. In this paper, we achieve high-quality image denoising by reducing or even eliminating this gap. Our method requires no training data or prior knowledge of the noise distribution. It consists of two lightweight networks that can be trained using only a single noisy test image. Specifically, we propose a random mask-based downsampler that generates multiple pairs of downsampled noisy images, which are similar but distinct. These image pairs serve as the input for the first network, with the mean image of each pair used as the target. This initially reduces the gap between the clean images of the input and target. Particularly, in our method, the clean counterpart of the first network's target (i.e., the mean image) can be obtained. We then train a second network using the mean image as input and its clean counterpart as the target. This effectively eliminates the gap and achieves better denoising results. Extensive experiments demonstrate that our method outperforms in both denoising performance and efficiency.

Infrared and visible image fusion (IVIF) aims to generate fused images with prominent targets and rich scene information. However, in low-light conditions, visible images lose accurate texture and color, reducing their ability to provide detailed scene information for fusion. Existing IVIF methods often overlook illumination degradation and cause color distortion when incorporating infrared information. To address these problems, we propose a novel visually enhanced IVIF method tailored for low-light environments. Our method combines low-light image enhancement (LLIE) and IVIF into a single module. First, we adaptively enhance the low-light visible image, ensuring rich texture and color for fusion. Additionally, we introduce a three-channel fusion coefficient map to transform infrared information into visible image, preventing color distortion and highlighting key targets while maintaining details of the fused image. Since infrared and visible images are from different modalities, we map them into the same high-dimensional feature space. We then propose the feature difference to integrate complementary information, producing a fused image with complete content and no redundancy. Notably, our method is zero-shot, requiring only a pair of test infrared and visible images for training. This better meets the complexity of IVIF in various low-light scenes. Extensive experiments show that in low-light conditions, our method surpasses other state-of-the-art (SOTA) methods by providing more natural colors, richer textures, and better alignment with human visual perception.

ABSTRACT Sparse Matrix‐Vector Multiplication (SpMV) is a fundamental operation in scientific computing, machine learning, and data analysis. The performance of SpMV on GPUs is crucial for accelerating various applications. However, the efficiency of SpMV on GPUs is significantly affected by irregular memory access patterns, high memory bandwidth requirements, and insufficient exploitation of parallelism. In this paper, we propose a Recursive Hybrid Compression (RHC) method to address these challenges. RHC begins by splitting the initial matrix into two portions: an Ellpack (ELL) portion and a Coordinate (COO) portion. This partitioning is followed by further recursive division of the COO portion into additional ELL and COO portions, continuing this process until predefined termination criteria, based on a percentage threshold of the number of nonzero elements, are met. Additionally, we introduce a dynamic partitioning method to determine the optimal threshold for partitioning the matrix into ELL and COO portions based on the distribution of nonzero elements and the memory footprint. We develop the RHC algorithm to fully exploit the advantages of the ELL kernel on GPUs and achieve high thread‐level parallelism. We evaluated our proposed method on two different NVIDIA GPUs: the GeForce RTX 2080 Ti and the A100, using a set of sparse matrices from the SuiteSparse Matrix Collection. We compare RHC with NVIDIA's cuSPARSE library and three state‐of‐the‐art methods: SELLP, MergeBase, and BalanceCSR. RHC achieves average speedups of 2.13, 1.13, 1.87, and 1.27 over cuSPARSE, SELLP, MergeBase, and BalanceCSR, respectively.

With the increasing expansion and complexity of cyberspace data and traffic, network security threats have also increased sharply. As one of the important means to ensure the security of information systems, intrusion detection is facing unprecedented challenges. In this paper, we propose a federated deep transfer learning algorithm, transfer-enhanced deep and transfer domain adaptation (TEDTDA), for intrusion detection. TEDTDA uses federated learning to train local models using intrusion detection data from multiple organizations to protect data privacy. It improves the efficiency of model training by integrating transfer learning theory and knowledge transfers. Moreover, it eliminates unreliable and low-quality local models through model selection in the training process to improve the detection effect. The algorithm is tested on three intrusion detection datasets: ISCX2012, NSL-KDD, and CICIDS2017. Compared with the benchmark algorithm, the proposed TEDTDA algorithm significantly improves the detection accuracy, training efficiency, and other key performance indicators.

Seafloor classification and mapping are crucial for understanding the composition and spatial distribution of submarine materials. Currently, data-driven modeling of underwater acoustic remote sensing data is widely adopted in seafloor classification. However, as research and applications expand, limitations of data-driven approaches—such as the lack of theoretical support, weak interpretability, and poor transferability—have become evident. In contrast, inherent geoacoustic properties of the seafloor, such as roughness and acoustic impedance, can be quantitatively inverted using acoustic scattering models. These properties exhibit objectivity, determinism, and interpretability, thereby providing robust constraints for data-driven modeling. Consequently, a novel seafloor classification strategy is proposed that integrates acoustic inversion with data-driven modeling. Firstly, a geoacoustic inversion model based on the angular response (AR) curve is established to obtain seafloor roughness and acoustic impedance as property features. Then, these features are fused with data features derived from acoustic backscatter mosaics and seafloor digital elevation models (DEMs). Finally, a deep neural network (DNN) is constructed for training, testing, and classification. To evaluate performance, the proposed method was applied to three survey areas with varying seafloor terrain complexity and validated by field data (seafloor photographs). Compared to purely data-driven approaches, integrating property features improved the overall classification accuracy by 4.6%, 3.9%, and 5.4%, respectively. Experimental results demonstrate that the proposed method effectively exploits seafloor geoacoustic properties, exhibiting strong generalization and transferability across diverse marine environments. This study provides a valuable reference for integrating acoustics theory into seafloor classification methods.

Mid-ocean ridges, seamounts, and back-arc basins are focal points for deep-sea research due to their abundant mineral and biological resources. Unlike the sediment-covered deep-sea plains, the above areas have complex topography due to frequent geological activity, which has greatly constrained the accuracy of seafloor classification using shipborne underwater acoustic remote sensing. With the development of autonomous underwater vehicles (AUVs), near range measurements become feasible. Here, we present a method for fusing acoustic and magnetic AUV data on a meter scale, thereby providing evidence for improved seafloor classification. First, a space scale normalized model was built for obtaining high spatial resolution seafloor magnetic anomalies. In the second step, nine features were extracted from the acoustic backscatter mosaic, the bathymetry, and the seafloor magnetic anomaly. Finally, deep neural network (DNN) models were built for training, testing, and classification. To evaluate the classification performance of the model, the method was applied to the survey of the Chinese contract area for polymetallic sulfide exploration on the Southwest Indian Ridge (SWIR) and validated by field data (seafloor video). The integration of acoustic and magnetic data, as opposed to using single acoustic data, improved the overall classification accuracy and kappa coefficient of DNN for basalts, breccias, and sediment by 6.4% and 0.096%, respectively. The experimental results show that the method can fully mine the acoustic and magnetic properties of the seafloor, effectively respond to the challenges of seafloor classification presented by the deep-sea complex environment, and provide a fresh idea for the research of near-bottom high-precision seafloor classification methods.

Signature estimation is critical for source evaluation and data processing. Far-field measurement is an accurate way to deterministically finalize the process and more importantly to clarify the source directivity. Based on the position-fixed mode, a seismic survey is carried out in India Ocean in 2021 with a deep-towed Janus-Helmholtz (JH) transducer as the source. This work shows a full view of acquiring, processing, and analyzing the far-field source signature. According to the source-receiver reciprocity theorem, an incidence angle of 7.4°–87.8° is covered in this survey. In order to overcome the positioning uncertainty of the ultrashort baseline mode, a novel joint ray-tracing inversion approach is presented on the basis of the conductivity-temperature-depth (CTD)-derived sound velocity model. On the data processing side, a customized workflow is established to attenuate the complex sea-floor reflections. It is realized based on sparse f–k transformation and adaptive subtraction and is validated effective from both the shot gather and spectrum panel. Analysis indicates that the signature of the transducer is jointly affected by the input assignment and the system response. It is potential to tune the two factors to optimize the signature, e.g., widening its bandwidth. Moreover, strong directivity is observed in terms of waveform consistency, amplitude decay, and energy distribution, which challenges the point-source propagation theory. Missing of the directivity information may cause severe imaging issues as illustrated by the comparison between global and directional deconvolution. This work could be a good guideline for future signature estimation and so-based data processing and analysis.

Abstract Top-K and selection operations are critical in data processing and analysis, and their efficient implementation on GPUs is increasingly important due to the growing demands of data analysis. Existing methods, primarily relying on the bucket partition execution model, encounter challenges such as uneven bucket distribution and latency in merging processes. To address these issues, we introduce a novel Split-Bucket Partition (SBP) execution model that specifically addresses these challenges. Additionally, we propose task and control flow optimizations targeted at top-K and selection algorithms, which further contribute to performance improvements. Our optimized algorithms significantly outperform existing approaches, delivering performance gains of up to $$2.3$$ 2.3 times and $$1.6$$ 1.6 times for different bucket partitioning rules. Our algorithms show robust performance improvements in non-uniform data scenarios, with gains ranging from $$1.9$$ 1.9 times to $$15.5$$ 15.5 times. However, it should be noted that the SBP model has limitations related to shared memory and register utilization, potentially impacting performance. Tests on TU102 and A100 GPU architectures validate the effectiveness of our approach, achieving a maximum speedup of $$2.9$$ 2.9 times. The study suggests that while the SBP model is effective for top-K and selection algorithms, it also holds promise for other computational tasks, setting the stage for future research.

Abstract Sparse matrix-vector multiplication (SpMV) plays a critical role in a wide range of linear algebra computations, particularly in scientific and engineering disciplines. However, the irregular memory access patterns, extensive memory usage, high bandwidth requirements, and underutilization of parallelism hinder the computational efficiency of SpMV on GPUs. In this paper, we propose a novel approach called block-wise dynamic mixed-precision (BDMP) to address these challenges. Our methodology involves partitioning the original matrix into uniformly sized blocks, with each block’s size determined by considering architectural characteristics and accuracy requirements. Additionally, we dynamically assign precision to each block using a precision selection method that takes into account the value distribution of the original sparse matrix. We develop two distinct SpMV computation algorithms for BDMP: BDMP-PBP (Precision-based partitioning) and BDMP-TCKI (Tailored compression and kernel implementation). BDMP-PBP partitions the matrix into two independent matrices for separate computations based on block precision, offering flexibility for integration with other optimization techniques. Meanwhile, BDMP-TCKI focuses on achieving significant thread-level parallelism and memory utilization by tailoring an appropriate compressed storage format and kernel implementation for each block. We compare BDMP with NVIDIA’s cuSPARSE library and three state-of-the-art SpMV methods, including SELLP, MergeBase, and BalanceCSR, using matrices from the University of Florida’s SuiteSparse dataset collection. BDMP-PBP and BDMP-TCKI show average speedups up to 2.64 $$\times $$ × and 2.91 $$\times $$ × on Turing RTX 2080Ti, and up to 2.99 $$\times $$ × and 3.22 $$\times $$ × on Ampere A100. The results demonstrate that BDMP enables the optimization of computation speed without compromising the precision necessary for reliable results.

Cloud storage provides an efficient and convenient way to manage data, but it also poses significant challenges to data security. The central issue with cloud storage is to ensure the ability of the data owner to control and manage the outsourced data. The password-protected secret sharing (PPSS) integrates password authentication and secret sharing to offer a fresh approach to secure private data. Users can share the risk of device corruption with a well-designed PPSS scheme and manage outsourced data with only human-memorizable passwords. To the best of our knowledge, none of the existing PPSS schemes can resist security threats in the post-quantum era, and there is an urgent need to design quantum-resistant solutions. However, post-quantum cryptography varies significantly from traditional cryptography, and it is challenging to design a quantum-resistant password-protected secret-sharing scheme for cloud storage. In this work, we take the first substantial step towards this challenge by proposing QPause, a quantum-resistant password-protected data outsourcing scheme for cloud storage. We first design a basic quantum-resistant PPSS scheme based on the lattice secure against semi-honest adversaries with a secure channel. On this foundation, we propose a quantum-resistant round-optimal password-protected data outsourcing scheme against strong adversaries. In addition, we formally prove that our scheme is secure and robust under various attacks against adversaries with quantum computing capabilities. The comparison results show that our new scheme outperforms its foremost counterparts.

Abstract Underwater images intuitively reflect the underwater environment information. However, underwater images have defects such as colour distortion and low contrast, which seriously affect the processing of complex underwater visual tasks. Here, a novel mixed model called mixed underwater image generative adversarial network (MUGAN) is presented, consisting of a generator and corresponding discriminator. The generator is built following the U‐shaped architecture, where a mixed block of convolution and self‐attention is developed. It effectively exploits the complementarity between the two paradigms. In addition, a dual discriminator is employed to induce the generator to produce realistic images at both the global semantic and local detail levels, which is only discriminates based on the patch‐level information. Meanwhile, a multi‐term loss function is formulated to supervise adversarial training by evaluating the perceptual quality of an image based on its global content, local texture and illumination smoothness. To validate the proposed approach, extensive experiments are conducted on the public underwater datasets. MUGAN achieves promising performance in terms of colour, contrast and naturalness, showing a significant improvement over other competitive models in visual quality and quantitative metrics.

Analyzing trajectory data can provide people with a higher quality of life. However, publishing trajectory data directly will leak privacy. The authors propose a trajectory data publication method based on differential privacy (TDDP). TDDP method consists of two stages. In the location generalization stage, firstly, the locations at each timestamp are clustered into classes by k-means++ algorithm, and then the representative location of each class is selected by using the exponential mechanism. In the generalized trajectory data publication stage, the authors design a sampling mechanism to form the generalized trajectories. The locations are sampled from the representative locations under different timestamps to form the generalized trajectories. The TDDP method can avoid the generation of non-semantic representative locations and ensure that the generalized trajectories can resist filtering attacks. The experimental results show that the trajectory data released by TDDP method can achieve a good balance between privacy protection and data availability.

Highly accuracy classification and mapping of seafloor hydrothermal fields provides an important addition to research into the geological background and exploration of seafloor massive sulphides (SMS). Currently, acoustic remote sensing using multibeam sounding systems (MBES) is the primary means of achieving large-scale seafloor classification. However, the characteristics of complex topography, complex seafloor distribution and deep water depth of the hydrothermal fields make it difficult to obtain accurate high-resolution seafloor classification maps using shipboard MBES surveys. Here a seafloor classification strategy combining shipboard MBES and autonomous underwater vehicle (AUV) sidescan sonar data is proposed. First, a downscaling model is established, which downscales the MBES backscatter mosaic (12 kHz, 10 m resolution) to a resolution of 2 m. The second step performs feature extraction of the downscaled MBES backscatter mosaic (12 kHz, 2 m resolution), the AUV sidescan sonar backscatter mosaic (150 kHz, 2 m resolution), and a seafloor digital elevation model (2 m resolution). Finally, a deep neural network model was built for training and classification. To evaluate the classification performance of the model, the method was applied to the survey of the Duanqiao hydrothermal field. Results were verified using field data (deep-tow video). The overall root mean square error and coefficient of determination (R2) for the classification were 0.032 and 0.840, respectively. The experimental results show that the method can effectively meet the challenges to seafloor classification presented by complex seafloor distribution, and can obtain an accurate high-resolution seafloor classification map.

Wireless sensor network is the momentous part of the Internet of Things. Data aggregation technology is the most practical method to reduce the amount of communication among nodes. Adding a privacy protection mechanism to data aggregation is one of the important means for privacy protection and security in wireless sensor networks. Aiming at certain performance defects of the existing SMART (Slice-Mix-AggRegaTe) privacy protection algorithms, a fat tree-based data aggregation privacy protection algorithm is proposed in this paper, which is referred to as the FTSMART (Fat Tree Slice-Mix-AggRegaTe) algorithm. Concerning the innovative algorithm, the fat tree (FT) is introduced, the fat tree structure is adopted to optimize the data slicing scheme and the aggregation tree generation scheme, and the allocation of fixed time intervals is employed for nodes to reduce the transmission collision in the data aggregation process and to guarantee the completion of the data transmission. The simulation experiment results demonstrate that the FTSMART algorithm has presented the favorable performance in terms of the privacy protection, the network communication overhead, and the data aggregation accuracy.

Since its introduction, IoT (Internet of Things) has enjoyed vigorous support from governments and research institutions around the world, and remarkable achievements have been obtained. The perception layer of IoT plays an important role as a link between the IoT and the real world; the security has become a bottleneck restricting the further development of IoT. The perception layer is a self-organizing network system consisting of various resource-constrained sensor nodes through wireless communication. Accordingly, the costly encryption mechanism cannot be applied to the perception layer. In this paper, a novel lightweight data integrity protection scheme based on fragile watermark is proposed to solve the contradiction between the security and restricted resource of perception layer. To improve the security, we design a position random watermark (PRW) strategy to calculate the embedding position by temporal dynamics of sensing data. The digital watermark is generated by one-way hash function SHA-1 before embedding to the dynamic computed position. In this way, the security vulnerabilities introduced by fixed embedding position can not only be solved effectively, but also achieve zero disturbance to the data. The security analysis and simulation results show that the proposed scheme can effectively ensure the integrity of the data at low cost.