Search

Abstract In low-light environments, the amount of light captured by the camera sensor is reduced, resulting in lower image brightness. This makes it difficult to recognize or completely lose details in the image, which affects subsequent processing of low-light images. Low-light image enhancement methods can increase image brightness while better-restoring color and detail information. A generative adversarial network is proposed for low-quality image enhancement to improve the quality of low-light images. This network consists of a generative network and an adversarial network. In the generative network, a multi-scale feature extraction module, which consists of dilated convolutions, regular convolutions, max pooling, and average pooling, is designed. This module can extract low-light image features from multiple scales, thereby obtaining richer feature information. Secondly, an illumination attention module is designed to reduce the interference of redundant features. This module assigns greater weight to important illumination features, enabling the network to extract illumination features more effectively. Finally, an encoder-decoder generative network is designed. It uses the multi-scale feature extraction module, illumination attention module, and other conventional modules to enhance low-light images and improve quality. Regarding the adversarial network, a dual-discriminator structure is designed. This network has a global adversarial network and a local adversarial network. They determine if the input image is actual or generated from global and local features, enhancing the performance of the generator network. Additionally, an improved loss function is proposed by introducing color loss and perceptual loss into the conventional loss function. It can better measure the color loss between the generated image and a normally illuminated image, thus reducing color distortion during the enhancement process. The proposed method, along with other methods, is tested using both synthesized and real low-light images. Experimental results show that, compared to other methods, the images enhanced by the proposed method are closer to normally illuminated images for synthetic low-light images. For real low-light images, the images enhanced by the proposed method retain more details, are more apparent, and exhibit higher performance metrics. Overall, compared to other methods, the proposed method demonstrates better image enhancement capabilities for both synthetic and real low-light images.

Abstract Remote sensing is gradually playing an important role in the detection of ground information. However, the quality of remote-sensing images has always suffered from unexpected natural conditions, such as intense haze phenomenon. Recently, convolutional neural networks (CNNs) have been applied to deal with dehazing problems, and some important findings have been obtained. Unfortunately, the performance of these classical CNN-based methods still needs further enhancement owing to their limited feature extraction capability. As a critical branch of CNNs, the generative adversarial network (GAN), composed of a generator and discriminator, has become a hot research topic and is considered a feasible approach to solving the dehazing problems. In this study, a novel dehazed generative adversarial network (GAN) is proposed to reconstruct the clean images from the hazy ones. For the generator network of the proposed GAN, the color and luminance feature extraction module and the high-frequency feature extraction module aim to extract multi-scale features and color space characteristics, which help the network to acquire texture, color, and luminance information. Meanwhile, a color loss function based on hue saturation value (HSV) is also proposed to enhance the performance in color recovery. For the discriminator network, a parallel structure is designed to enhance the extraction of texture and background information. Synthetic and real hazy images are used to check the performance of the proposed method. The experimental results demonstrate that the performance can significantly improve the image quality with a significant increment in peak-signal-to-noise ratio (PSNR). Compared with other popular methods, the dehazing results of the proposed method closely resemble haze-free images.

At present, the construction of digital government has entered a deepwater area.The government big data platform, as a data base, supports various government information applications.The security and compliance of its private data has been widely concerned by the industry.Federated learning is an important method to effectively solve data silos, and the application of government big data platforms based on federated learning has high research value.Firstly, the current status of government big data platforms and its federated learning application were introduced.Then this paper analyzed three major management challenges involved in the collection, classification and grading and sharing of privacy data on government big data platforms.Further, the problem-solving methods of federated learning based recommendation algorithms and privacy intersection techniques were explored.Finally, summaries and prospects were made for the future application of privacy data on government big data platforms.

In order to improve the recognition rate of the tone classification of doctors in online medical services scenarios, we propose a model that integrates a one-dimensional convolutional neural network (1DCNN) with a bidirectional long short-term memory network (BiLSTM). Firstly, significant tone types within online medical services scenarios were identified through a survey questionnaire. Secondly, 68 features in both the time and frequency domains of doctors’ tone were extracted using Librosa, serving as the initial input for the model. We utilize the 1DCNN branch to extract local features in the time and frequency domains, while the BiLSTM branch captures the global sequential features of the audio, and a feature-level fusion is performed to enhance tone classification effectiveness. When applied in online medical services scenarios, experimental results show that the model achieved an average recognition rate of 84.4% and an F1 score of 84.4%, significantly outperforming other models and effectively improving the efficiency of doctor-patient communication. Additionally, a series of ablation experiments were conducted to validate the effectiveness of the 1DCNN and BiLSTM modules and the parameter settings.

The traditional methods for identifying sensitive data in APIs mainly encompass rule-based and machine learning-based approaches. However, these methods suffer from inadequacies in terms of security and robustness, exhibit high false positive rates, and struggle to cope with evolving threat landscapes. This paper proposes a method for detecting sensitive data in APIs based on the Federated Large Language Model (FedAPILLM). This method applies the large language model Qwen2.5 and the LoRA instruction tuning technique within the framework of federated learning (FL) to the field of data security. Under the premise of protecting data privacy, a domain-specific corpus and knowledge base are constructed for pre-training and fine-tuning, resulting in a large language model specifically designed for identifying sensitive data in APIs. This paper conducts comparative experiments involving Llama3 8B, Llama3.1 8B, and Qwen2.5 14B. The results demonstrate that Qwen2.5 14B can achieve similar or better performance levels compared to the Llama3.1 8B model with fewer training iterations.

Federated learning (FL) demonstrates significant potential in Industrial Internet of Things (IIoT) settings, as it allows multiple institutions to jointly construct a shared learning model by exchanging model parameters or gradient updates without the need to transmit raw data. However, FL faces risks related to data poisoning and model poisoning. To address these issues, we propose an efficient verifiable federated learning (EVFL) method, which integrates adaptive gradient sparsification (AdaGS), Boneh–Lynn–Shacham (BLS) signatures, and fully homomorphic encryption (FHE). The combination of BLS signatures and the AdaGS algorithm is used to build a secure aggregation protocol. These protocols verify the integrity of parameters uploaded by industrial agents and the consistency of the server’s aggregation results. Simulation experiments demonstrate that the AdaGS algorithm significantly reduces verification overhead through parameter sparsification and reuse. Our proposed algorithm achieves better verification efficiency compared to existing solutions.

As an important component of inlet and exhaust mufflers, the acoustic characteristics of perforated components are inevitably affected by the flow of air. Therefore, obtaining the acoustic impedance of the perforated element under airflow conditions is a prerequisite for accurate calculation of the muffler’s muffling performance. In this work, the frequency domain linear Navier–Stokes (L-NS) method is used to extract the acoustic impedance of perforated plates under grazing flow. The predicted perforated acoustic impedance is consistent with the calculation results of published acoustic impedance expression, and the impedance boundary condition is defined to calculate the transmission loss (TL) of the perforated muffler, which agrees well with the experimental results and verifies the accuracy of the method. The effect of perforation angles on the transmission loss of mufflers in different Mach numbers ( Ma ) and aperture plate thickness ratio ( d h / t p ) is analyzed by the frequency domain L-NS method. The results show that when 1≤ d h / t p

Federated learning is a distributed learning method used to solve data silos and privacy protection in machine learning, aiming to train global models together via multiple clients without sharing data. However, federated learning itself introduces certain security threats, which pose significant challenges in its practical applications. This article focuses on the common security risks of data poisoning during the training phase of federated learning clients. First, the definition of federated learning, attack types, data poisoning methods, privacy protection technology and data security situational awareness are summarized. Secondly, the system architecture fragility, communication efficiency shortcomings, computing resource consumption and situation prediction robustness of federated learning are analyzed, and related issues that affect the detection of data poisoning attacks are pointed out. Thirdly, a review is provided from the aspects of building a trusted federation, optimizing communication efficiency, improving computing power technology and personalized the federation. Finally, the research hotspots of the federated learning data poisoning attack situation prediction are prospected.

In a modern electric power transmission grid, the phasor measurement unit requires a reliable transport of its sampled statistics with a low end-to-end failure rate (EEFR) to ensure the accuracy of the grid state estimation. However, EEFR can be deteriorated by packet losses due to multiple link disruptions in the primary forwarding path (PP). To address that, we investigate a novel disruption resilient transport protocol (DRTP) enabling hop-by-hop retransmission utilizing the redundant subpaths (RSPs) available for the PP to increase reliability. It addresses the new distributed collaboration issue under multiple link failures to avoid cache mismatching. These have not been considered by the existing approaches. The DRTP was evaluated in the ndnSIM simulator through both the typical and general routes that are constructed from real transmission grids. The numerical results demonstrate that it has a significant advantage in reducing the EEFR with a low end-to-end delivery time under serious link disruptions.

The importance of network security has become increasingly prominent due to the rapid development of network technology. Network intrusion detection systems (NIDSs) play a crucial role in safeguarding networks from malicious attacks and intrusions. However, the issue of class imbalance in the dataset presents a significant challenge to NIDSs. In order to address this concern, this paper proposes a new NIDS called CBF-IDS, which combines convolutional neural networks (CNNs) and bidirectional long short-term memory networks (BiLSTMs) while employing the focal loss function. By utilizing CBF-IDS, spatial and temporal features can be extracted from network traffic. Moreover, during model training, CBF-IDS applies the focal loss function to give more weight to minority class samples, thereby mitigating the impact of class imbalance on model performance. In order to evaluate the effectiveness of CBF-IDS, experiments were conducted on three benchmark datasets: NSL-KDD, UNSW-NB15, and CIC-IDS2017. The experimental results demonstrate that CBF-IDS outperforms other classification models, achieving superior detection performance.

The development of domestic high resolution remote sensing satellite technology provides a new means for the investigation of macroscopic, local and detailed characteristics of the Quaternary.Alluvial-diluvial fans, valley terraces, glacial deposits are widely distributed in the eastern section of China-Pakistan Economic Corridor. In order to study the applicability of domestic high-resolution data in the Quaternary geological survey of this region, the alluvial and flood fan in front of Uruat mountain along Gaizi River basin of China-Pakistan highway and three typical Quaternary strata of Gaizi River valley terrace and Kraguirake glacial moraine were selected as the research objects in this study.Using domestic GF-2 images as the main data sources, with the aid of digital elevation model data, the remote sensing interpretation marker was established, and the fine geological interpretation was carried out.The composition and variation were found out through field verification.The previous distribution range of different periods of glacial moraines in Kailaillak was revised.Based on the analysis of the terrace in Gaizi River valley, the study area experienced at least five stages of tectonic uplift since late Epipleistocene, and the terrace base was formed by at least two stages of glaciation.This study shows that GF-2 images can not only quickly identify the change characteristics of landforms and loose deposits at macro scale, but also it can get geological phenomena that cannot be observed by conventional methods.And then it can meet the requirements of large scale interpretation and cartography.Especially in the identification of micro-geomorphology and interpretation of Quaternary strata, it can improve the level of fine geological interpretation.Research results can provide basic geological data for understanding the development and evolution process of fluvial and glaciation in the Gaizi River basin.Besides, this study can provide a typical example for remote sensing investigation of Quaternary soil mass along the China-Pakistan highway.

The dynamic migration technology of virtual machine provides a strong support for the resource scheduling of virtualization system.As one of the two core algorithms of virtual machine dynamic migration, Post-Copy algorithm has been a hot issue for scholars at home and abroad for its advantages of stable migration time and short migration downtime.The fault tolerance mechanism of virtual machine, the connection between the transfer mode of memory page and the page missing error during the migration process, and the source code of QEMU-KVM platform were studied.A fault tolerant method based on transaction synchronization was proposed to improve the stability of the Post-Copy migration algorithm.The experimental results show that the proposed algorithm can ensure the fast repair of deep end virtual machine failure, target virtual machine failure and network fault during the migration process, and solve the stability problem at a small cost.The method proposed effectively improves the stability of the Post-Copy migration algorithm, and provides reference for the future optimization research direction.

Adversarial examples (AEs) against deep neural networks (DNNs) raise wide concerns about the robustness of DNNs. Existing detection mechanisms are often limited to a given attack algorithm. Therefore, it is highly desirable to develop a robust detection approach that remains effective for a large group of attack algorithms. In addition, most of the existing defences only perform well for small images (e.g. MNIST and Canadian institute for advanced research (CIFAR)) rather than large images (e.g. ImageNet). In this paper, the authors propose a robust and effective defence method for analysing the sensitivity of various AEs, especially in a much harder case (large images). Their method first creates a feature map from the input space to the new feature space, by utilising 19 different feature mapping methods. Then, a detector is learned with the machine‐learning algorithm to recognise the unique distribution of AEs. Their extensive evaluations on their proposed detector show that their detector can achieve: (i) low false‐positive rate (

Rootkit is a set of persistent and undetectable attack technologies,which can hide their attack behavior and backdoor trace by modifying software or kernel in operating system and changing execution path of instruction.Firstly,the basic definition and evolution of Rootkit were introduced,then the operating principle,current mainstream technology and detection methods of Rootkit were discussed.Then,through comparative experiments on performance and security,the application of mimic defense system was described for Web based on dynamic,heterogeneous,redundant structure under Trojan Horse attack.Experiments show that mimic defense system can effectively defend against Trojan Horse in tests in the premise of low overhead.At last,the opportunities and challenges of the DHR system were summarized.

The unbalanced development status of network security was introduced.The main hazards and the mechanism model of penetration testing were described,and the inherent shortcomings of many existing traditional defense methods were analyzed.However,new method of the mimic defense model makes the attack information obtained invalid by dynamically selecting the executive set and adaptively changing the system composition.The same attack mode is difficult to be maintained or reproduced.Based on the attack chain model,the traditional defensetechnology and mimic defense technology were analyzed and compared,and it was demonstrated that it had a protective role in multiple stages of the attack chain.Finally,the effectiveness and superiority of the mimic defense was verified by experiments,and the model was summarized and prospected.

Software-defined networking (SDN) enables dynamic management and flexible network control by employing reactive rule installation. Due to high power consumption and cost, current OpenFlow switches only support a limited number of flow rules, which is a major limitation for deploying massive fine-grained policies. This bottleneck can be exploited by attackers to launch saturation attacks to overflow the flow table. Moreover, flow table overflow can occur in the absence of malicious attackers. To cope with this, researchers have developed many proposals to relieve the load under benign conditions. Among them, the dynamic timeout mechanism is one of the most effective solutions. We notice that when the SDN controller adopts dynamic timeouts, existing flow table saturation attacks can fail, or even expose the attackers, due to inaccurate inferring results. In this paper, we extract the common features of dynamic timeout strategies and propose an advanced flow table saturation attack. We explore the definition of flow rule lifetime and use a timing-based side-channel to infer the timeout of flow rules. Moreover, we leverage the dynamic timeout mechanisms to proactively interfere with the decision of timeout values and perform an attack. We conduct extensive experiments in various settings to demonstrate its effectiveness. We also notice that some replacement strategies work differently when the controller assigns dynamic timeouts. The experiment results show that the attack can incur significant network performance degradation and carry out the attack in a stealthy manner.

Rolling bearing fault diagnosis is one of the challenging tasks and hot research topics in the condition monitoring and fault diagnosis of rotating machinery. However, in practical engineering applications, the working conditions of rotating machinery are various, and it is difficult to extract the effective features of early fault due to the vibration signal accompanied by high background noise pollution, and there are only a small number of fault samples for fault diagnosis, which leads to the significant decline of diagnostic performance. In order to solve above problems, by combining Auxiliary Classifier Generative Adversarial Network (ACGAN) and Stacked Denoising Auto Encoder (SDAE), a novel method is proposed for fault diagnosis. Among them, during the process of training the ACGAN-SDAE, the generator and discriminator are alternately optimized through the adversarial learning mechanism, which makes the model have significant diagnostic accuracy and generalization ability. The experimental results show that our proposed ACGAN-SDAE can maintain a high diagnosis accuracy under small fault samples, and have the best adaptation performance across different load domains and better anti-noise performance.

The reliability of the insulator has directly affected the stable operation of electric power system. The detection of defective insulators has always been an important issue in smart grid systems. However, the traditional transmission line detection method has low accuracy and poor real-time performance. We present an insulator defect detection method based on CenterNet. In order to improve detection efficiency, we simplified the backbone network. In addition, an attention mechanism is utilized to suppress useless information and improve the accuracy of network detection. In image preprocessing, the blurring of some detected images results in the samples being discarded, so we use super-resolution reconstruction algorithm to reconstruct the blurred images to enhance the dataset. The results show that the AP of the proposed method reaches 96.16% and the reasoning speed reaches 30FPS under the test condition of NVIDIA GTX 1080 test conditions. Compared with Faster R-CNN, YOLOV3, RetinaNet and FSAF, the detection accuracy of proposed method is greatly improved, which fully proves the effectiveness of the proposed method.

Given users and products that he/she reviews, can we recognize fake reviews just using the text information, or determine whether a reviewer is a fraud or not? Automatically detecting fake reviews and reviewers is an urgent problem and lots of work attempts for discovering linguistics, behaviors and graph patterns. However, in reality, there are new kinds of fraudsters who can change their behaviors to camouflage as genuine reviewers to avoid detection systems. With the fraudsters become distributed, dynamic, and adversarial, anti-spam tasks face a new challenge. In this paper, we tackle the challenge of adversarial fraudsters in online app review platform and propose a system called DDF (Detect, Defense, and Forecast) to uncover camouflage accounts. Firstly, we select a small set of seed with high-precision based on text and behavior features; Secondly, we build our graph-based detection model for uncovering hidden (distant) users who serve structurally similar to the seed by utilizing Graph Convolutional Network (GCN) algorithm. Thirdly, we evaluate DDF using real-world data set from Tencent APP Store and analyze the potential fraudsters detected by DDF. It is worth mentioning that precision can achieve 0.95+ . Finally, we validate the efficiency and scalability of DDF and show that it can be well transferred to other anti-spam tasks.

Network function virtualization allows network functions to be implemented and managed flexibly as a service function chain (SFC) in the data plane to process flows. However, software-based SFCs lead to poor performance compared to proprietary middleboxes. Moreover, existing solutions tackling performance issues suffer from the development complexity incurred by hardware details. To address these problems, we leverage both the high performance of P4-capable devices and the high flexibility of P4 language. In this paper, we present a P4 Service Chaining framework (P4SC), which tackles multiple challenges for P4 to support the SFC implementation. P4SC provides a suite of primitives allowing efficient SFC expression, and a converter and a generator converting input SFC requests to the corresponding P4 program. Here, an algorithm based on longest common subsequence (LCS) is used to allow simultaneously implementing multiple SFCs. Moreover, P4SC offers a runtime manager for flexible SFC management at runtime. It also provides an automatic integration mechanism to integrate P4-based NFs into P4SC. We implement a P4SC prototype, which supports three types of P4-capable devices. The experimental results show that P4SC outperforms state of the arts with orders-of-magnitude SFC performance improvement while maintains high flexibility.

To solve the problem that the details of fusion images are not retained well and the information of feature targets is incomplete, we proposed a new fusion method of infrared (IR) and visible (VI) image—IR and VI image fusion method of dual non-subsampled contourlet transform (NSCT) and pulse-coupled neural network (PCNN). The method makes full use of the flexible multi-resolution and multi-directional of NSCT, and the global coupling and pulse synchronization excitation characteristics of PCNN, effectively combining the features of IR image with the texture details of VI image. Experimental results show that the algorithm can combine IR and VI image features well. At the same time, the obtained fusion image can better display the texture information of image. The fusion performance in contrast, detail information and other aspects is better than the classical fusion algorithm, which has better visual effect and evaluation index.

The isothermal bulk moduli of anhydrous Mg2SiO4-ringwoodite (Rw) and Fe2SiO4-Rw, and other 4–2 oxide spinels at ambient P-T condition have been evaluated, and empirically fitted to a model as KT0 = 270.8(300) + 0.343(59)*V0 + 23.04(269)*EN-total, where KT0 is the isothermal bulk modulus in GPa, V0 the unit-cell volume in Å3 and EN-total the electronegativity total of all cations in the chemical formula. This model well reproduces all data used in its calibration, and may be used to predict the KT0 of other 4–2 oxide spinels. Combined with the generally linear volume–composition relationship of the Rw solid solutions along the join Mg2SiO4–Fe2SiO4, this model leads to a much smaller composition effect on the KT0: KT0 = 185.0(1) + 7.0(1)*XFe, where XFe is the atomic ratio Fe/(Fe + Mg). Furthermore, a bulk composition-independent compositional variation with P has been disclosed for the Rw at the P-T conditions of the lower part of the mantle transition zone (MTZ): XFe = 0.222(41) – 0.0053(19)*P, with P in GPa. The nearly ideal mixing behavior, much smaller composition effect on the bulk modulus, and significant compositional variation of the Rw in the lower part of the MTZ substantially increase the gradients of the Vs-P and Vp-P profiles to generally match those constrained by the seismic reference models PREM and AK135. If there is any global low-T anomaly at the depth of 660 km, its required magnitude is most likely not larger than 200 K.

The approximation algorithm of single virtual network mapping solution aimed to minimize the mapping cost based on convex quadratic programming relaxation was designed. Then aiming for dynamic arrival construction request of the single virtual network，the mapping scheme was achieved based on physical network resource pricing strategy by shadow price，applying above-mentioned approximation algorithm. And then completed the competitive algorithm design based on virtual network admission control strategy by mapping cost constraint and provided the competitive analysis of this algorithm. Experiment results show that the proposed algorithm increases the effective utilization of physical network resources，hence it can improve the virtual network construction request acceptance ratio and the long-term profit of physical network service provider.

Firstly，the concept of SDN applications was proposed from the perspective of users payment for allocating network resources.Secondly，with the comparison between traditional desktop application programming and network programming，the advantage of SDN programming was described.Thirdly，the concept of software defined price （SDP）was introduced.Based on the characteristics of SDN applications，the framework of SDP-based SDN application so as to apply to the OpenDaylight open source project was designed.Finally，an implementation of creating a virtual tenant network based on SDP was showed.The results of experiment present that the users can choose their desired virtual tenant according to the mechanism of SDP.

To improve the feasibility of virtual node mapping，grounded on feasibility test theorem and node rank indicators used to measure node availability，the virtual network mapping iterative algorithm based on K-best perfect matchings of bipartite graph was proposed，aiming at reducing mapping cost.Experiment shows that the proposed algorithm increases virtual network construction request acceptance ratio and benefit cost ratio，and hence can improve profit of physical network service provider.

Analysis of crop breeding technology is one of the important means of computer-assisted breeding techniques which have huge data, high dimensions, and a lot of unstructured data. We propose a crop breeding data analysis platform on Spark. The platform consists of Hadoop distributed file system (HDFS) and cluster based on memory iterative components. With this cluster, we achieve crop breeding large data analysis tasks in parallel through API provided by Spark. By experiments and tests of Indica and Japonica rice traits, plant breeding analysis platform can significantly improve the breeding of big data analysis speed, reducing the workload of concurrent programming.

Comparative analysis whether considering the lateral inertia or not, aiming at the longitudinal vibration of the drill string in drilling progress, is proposed. In the light of the actual condition, the mechanical model of the drill string about vibration is established on the basis of the theoretical analysis. Longitudinal vibration equation of the drill string is derived from the Rayleigh-Love model and one-dimensional viscoelastic model. According to the Laplace transform method and the relationships among parameters of the model, the solutions to complex impedance at the bottom of the drill string are obtained, and then the comparison results are analyzed, which is the lateral inertia effect on longitudinal vibration characteristics. The researches show that the smaller the length of the drill string, the greater the cross-sectional area of the drill string, the greater the damping coefficient of bottom hole on the bottom of the drill string, and the more evident the effect on the dynamic stiffness of the drill string with lateral inertia effect. The Poisson ratio of the drill string only has some effects on it taking account of the lateral inertia effect, and the influence is relatively small compared with the former three conditions.