最新刊期
卷 71 ， 期 2 ， 2025
- “在人工智能领域，专家建立了深度学习体系，为智能技术发展提供新方向。”
- “在数字化生活领域，专家深入分析了Android多开应用的安全威胁，评估了角色间协同作用可能导致的风险，为解决多开应用安全问题提供解决方案。”摘要：The evolution of digital life has cemented Android multi-instance applications as a staple in users’ online interactions. By sharing user ID(UID) and employing app virtualization technology, these applications offer an installation-free experience, surpassing traditional permission limitations. However, this convenience exposes data privacy vulnerabilities and escalates security risks in multi-instance environments. The intricate interplay among different roles further complicates the threat landscape with increased diversity and subtlety. We initiated an exhaustive analysis of security threats within the multi-instance environment, assessing risks stemming from the synergy of multiple participant roles. Subsequently, we systematically reviewed the core identifying features of the virtualization framework and provided a comprehensive summary of current attack methods and defense strategies. We conclude by critically examining the research limitations and ventureing into prospective avenues for future investigation, aiming to confront the challenges inherent in app-virtualization-based Android cloning applications.关键词：mobile security;app-virtualization;multi-instance applications;security threats218|356|0更新时间：2025-05-09
- “深度学习技术在网络异常流量检测领域的研究进展，为维护网络安全提供新思路。”摘要：With the rapid growth of data volume and increasing diversity of traffic types, abnormal traffic detection in networks has become more complex. The effective detection of abnormal traffic is crucial for promptly identifying potential network threats and attacks, thereby maintaining the security of the network. This paper aims to provide a comprehensive review of the current state of research on deep learning-based network abnormal traffic detection, the challenges faced by current deep learning-based methods, and future research directions. First, this study elaborates on the fundamental concepts and detection process of abnormal network traffic detection, followed by an analysis of the commonly used datasets for this purpose. Subsequently, it delves into the advantages and limitations of various deep learning methods, including convolutional neural networks, bidirectional long short-term memory networks, autoencoders, generative adversarial networks, graph neural networks, and natural language processing techniques for abnormal traffic detection. Finally, the main challenges encountered by deep learning-based abnormal traffic detection methods are discussed, and future research directions are proposed.关键词：abnormal traffic detection;traffic analysis;network security;deep learning1206|259|0更新时间：2025-05-09
- “在深度伪造检测领域，研究者提出了音脸协同驱动的SFSD算法，通过多模态模型SFformer提升检测性能，为伪造视频识别提供新方案。”摘要：Most mainstream deepfake detection methods are based on visual single-modality, which detects deepfake videos by identifying fake artifacts in video frames. However, different deepfake methods may introduce different artifacts; therefore such methods have limited performance and poor generalization ability. Meanwhile, the current work utilizing audio information does not fully leverage audio information to uncover tampering in the visual modality. Natural videos have intrinsic synergy between speeches and faces, and deepfake methods this disrupt the intrinsic synergy between speeches and faces; this paper proposes a speech-face synergy-driven deepfake detection algorithm, called SFSD (Speech-Face Synergy Detection). The core of the algorithm is the audio-face synergy contrastive learning strategy, which constructs samples on a general video dataset to simulate the disruption of audio-face synergy using forgery methods and pre-trains the detection model on these samples. This strategy utilizes a large number of unlabeled real videos and enhances model performance and generalization capability. A multimodal model named SFformer (Speech-Face Transformer) is constructed, which utilizes an attention bottleneck to guide the condensation and fusion of essential information in the audio-face modality, reduces interference from redundant information, improves the model's feature extraction capability, and enhances detection performance. Numerous experiments on the public dataset FakeAVCeleb demonstrate that the accuracy of SFSD can reach 72.12% after pre-training, surpassing some benchmark methods, and the accuracy reaches 89.51% after transfer learning, which is higher than that of previous studies, and improves the generalization ability.关键词：deepfake detection;contrastive learning;multimodal fusion167|140|0更新时间：2025-05-09
- “在数字资产安全领域，专家提出了基于TrustZone的以太坊密钥保护方法，提高了用户密钥安全性，为以太坊系统安全运行提供了解决方案。”摘要：The security of Ethereum digital assets mainly depends on users’ private keys. If a user’s private key is leaked, thier digital assets are irretrievably lost. Cryptocurrency wallets typically provide key management and protection for blockchain systems. However, the security of most existing cryptocurrency wallets is poor. To provide better protection for users’ digital assets, we propose a TrustZone-based key protection scheme for the Ethereum blockchain, which can significantly enhance the security of the generation, storage, and usage of Ethereum keys without disrupting the standard functionalities of Ethereum when users access and use the Ethereum system through mobile terminals that support TrustZone. Mainly, operations related to private keys are enforced in the secure world of the TrustZone. Private keys that are not in service are stored in the local filesystem as secure files. The corresponding secure file is loaded into the secure world to recover the private key only when a transaction needs to be signed, ensuring that private keys can be used securely and continuously. We implemented the prototype system in a simulation environment and verified its functionality and performance. The experimental results show that the developed prototype system can work normally in Ethereum with a reasonable performance overhead, validating the feasibility of the proposed scheme.关键词：Ethereum;cryptocurrency wallet;key protection;TrustZone;secure file184|139|0更新时间：2025-05-09
- “在基因组学领域，HMDA算法通过HMAC技术提升DNA组装的准确性、安全性和效率，为基因组研究提供新方案。”摘要：DNA assembly is a crucial step in genomic research. Traditional genome assembly algorithms based on second-generation sequencing technology have the following problems: 1) they cannot guarantee the integrity and accuracy of the data; 2) they cannot handle the problem of assembling mixed species genome sequences; 3) they consume a large amount of memory space; and 4) they lack security guarantees. A DNA assembly algorithm based on a Hash-based message authentication code(HMAC), HMDA, is proposed to address these issues. In the DNA assembly process, HMAC technology is used to encode assembly information, and the properties of codons are used to embed HMAC-encrypted information content into the genome. This information is extracted as a key basis for screening the correct sequence when performing the assembly operations. In addition, HMDA maps different keys to different species or users individually, enabling species-specific or user-specific authentication. Experimental results show that this algorithm has the following advantages over other assembly algorithms:1) it can generate accurate and complete assembly results for sequencing libraries that theoretically recover genomes ;2) it can assemble different species from mixed-species genomic data ;3) it utilizes at most 33.4% of the space resources required by other algorithms; 4) it enhances security by producing different assembly results for authorized and ordinary users. HMDA can detect data tampering and return error status codes.关键词：DNA assembly;genomics;second-generation sequencing technology;Hash-based message authentication code (HMAC)67|103|0更新时间：2025-05-09
- “在网络安全领域，专家设计了可聚合共享棘轮树算法，并提出了基于此的安全可追溯组密钥协商协议TGKA，有效解决了组密钥协商中的通信效率低、计算开销大等问题，为动态组的高效密钥更新提供了解决方案。”摘要：Group key agreement protocols are designed to address the security issues of group message transmission among multiple participants in an insecure communication network. However, existing solutions often suffer from low communication efficiency, high computational overhead, difficulty in scaling group sizes, and insufficient security issues. To address these problems in group key agreement, a novel aggregatable shared ratchet tree algorithm is designed. Based on this, a secure and traceable group key agreement(TGKA) protocol is proposed. The TGKA protocol combines a signature-based aggregatable broadcast scheme with the concept of key encapsulation. The protocol reduces computational and communication overhead by using a ratcheting tree to divide users into multiple subgroups and perform key agreements within these subgroups, achieving efficient key updates for dynamic groups. The experimental results indicate that while ensuring the security of the group key agreement protocol, the TGKA protocol can reduce the communication complexity for both group senders and receivers. This demonstrates a certain level of feasibility in medium-sized groups with tens to hundreds of users.关键词：group key agreement;forward safety and backward safety;traceability;ratchet tree98|261|0更新时间：2025-05-09
- “在司法领域，专家提出了基于本体提示指导的命名实体识别方法，通过构建知识图谱本体模型和指令微调大语言模型，有效提升了少样本条件下的识别效果。”摘要：To enhance the efficacy of complex judicial named entity recognition in low-resource and limited-sample scenarios, this study proposes a judicial named entity recognition method based on ontology prompt guidance, utilizing a large language model. First, taking counterfeit judicial documents as an example, a “top-down” approach is used to construct a knowledge graph ontology model of counterfeit judicial documents based on the document content and the existing ontology in the field of judicial documents. Then, the instruction is constructed based on the ontology model. The instruction includes four parts: task description, ontology description, task example and judicial text. A large language model is fine-tuned to complete the judicial named entity recognition task. The ontology description section adds entity definitions and relationship information to the fine-tuning process. Finally, we selected 12 types of fine-grained entities included in the ontology model constructed in this study and collected the datasets for experiments,and compared these results with those obtained using several typical traditional entity recognition methods. The results showed that the method proposed in this study performed better under the condition of small sample fine-tuning.关键词：judicial documents;named entity recognition;large language models;prompt engineering;ontology235|169|0更新时间：2025-05-09
- “医学指南知识抽取模型，提升了指南证据抽取效率和偏倚风险识别能力，推动高质量临床医学研究。”摘要：In recent years, the publication speed of medical guidelines has been rapid, but their structured organization and retrieval in the evidence-based process face many challenges, which affect the rapid acquisition and application of knowledge. This study proposes a guideline knowledge extraction model based on evidence-based medicine to address the issue. This model aims to efficiently extract guideline evidence and identify the risk of bias in the source literature of guideline evidence. The model combines manually annotated small-sample datasets to construct prompt instructions for training large language models to improve their performance in guideline knowledge extraction tasks. A randomized controlled trial bias risk assessment model was constructed that integrated external knowledge, and enhanced the model's bias risk identification ability by utilizing diverse external knowledge information based on the BERT model. After fine-tuning with prompts, the performance of the large language model in extracting key information from guidelines was significantly improved, with BLEU-4 improving by over 43 percentage points. The evidence bias risk assessment model that integrates external knowledge also showed superior performance in identifying bias risks, with a precision of 0.898 and an F1 value of 0.892. With a limited dataset size, fine-tuning a large language model with hint engineering can effectively extract the required medical guideline information. The incorporation of various external knowledge sources substantially improved the effectiveness of the bias risk identification model. This suggests that models designed to extract knowledge from medical guidelines can contribute to enhancing the quality of clinical research.关键词：evidence-based medicine;large language model;Bert;bias risk assessment132|156|0更新时间：2025-05-09
- “在云计算资源消耗预测领域，专家设计了高精度预测框架，通过自组织映射网络、注意力机制和改进遗传算法，验证了其在短期和长期预测上的优势，为提高预测精度提供了新方案。”摘要：The current accuracy of cloud computing resource consumption forecasts can be improved. This study introduces an innovative resource forecasting framework to address the challenges of predicting resource usage within cloud data centers. The framework includes data preprocessing, a long short-term memory network based on an attention mechanism, and a search mechanism based on an improved genetic algorithm. The intrinsic properties of the data are preserved by the self-organizing mapping network, which helps remove noise and reveal the underlying patterns of the data. The combination of the attention mechanism and long short-term memory network enhances the model's attention to key time steps, thus enhancing the prediction ability. An improved genetic algorithm was used for weight optimization so that the model could adaptively adjust the attention weight allocation. In the experiment, the framework proposed in this study was validated based on Google cluster tracking data. Based on a comparison with the benchmark algorithm, the proposed framework has advantages in the three indices for both short-term and long-term predictions, which proves that the framework has higher accuracy. The modified Friedman test and Nemenyi test proved that the proposed method’s prediction performance is better than that of the two benchmark algorithms in terms of the root-mean-square error.关键词：cloud data center;resource utilization;long short⁃term memory network;attention mechanism136|245|0更新时间：2025-05-09
-
Domain Information Self-Augmenting Extreme Random Tree Model Applied to Time Series Data Prediction AI导读
“在时序数据预测领域，专家提出了基于域信息自增极端随机树模型，通过时序平移器和特征分解模块优化预测精度，为降低风险和优化资源配置提供新方案。”摘要：Time-series data prediction is pivotal in addressing uncertainty, mitigating risks, optimizing resource allocation, and enhancing service quality. However, the complex nature of time-series data challenges accurate prediction. To address this issue, we propose a domain information self-augmenting extreme random tree model applied to time-series data prediction. We employ a time-series shifter to extract relationships among the data points, utilize a feature decomposition module to eliminate inherent noise from the original sequence, and leverage a generative adversarial network-based feature enhancement module to effectively capture static and spatiotemporal features of the time-series data. Furthermore, we optimized the parameters of our time prediction model. Taking public health emergencies as an example, we forecasted key indicators’ trends in different states of the United States. The experimental results demonstrate that our model performs well in predicting daily new cases during public health emergencies in Massachusetts, Maryland, and Washington State; however, it exhibits slight limitations when forecasting daily new deaths.关键词：feature decomposition;generative adversarial network;extreme random trees;domain information enhancement201|165|0更新时间：2025-05-09 - “在色玻璃凝聚有效场理论框架下，研究了质子形状涨落效应对EIC和LHeC能区的电子-核深度虚康普顿散射微分截面的影响。通过热点模型引入质子形状涨落效应，改进了偶极子散射振幅，从而获得一个质子形状涨落效应修正的深度虚康普顿散射微分截面。计算了深度虚康普顿散射微分截面，把没有涨落效应与包含涨落效应的截面作比较，发现在电子-质子深度虚康普顿散射过程中，质子形状涨落对小动量转移区间的截面有一定的影响，使得微分截面被压低；电子-核在深度虚康普顿散射过程中，质子形状涨落对相对较大的动量转移区间的截面有一定的影响，从而导致衍射图案的波峰和波谷的位置向右移动。衍射图案中波峰和波谷位置容易进行高精度测量，电子-核深度虚康普顿散射过程中波峰和波谷位置的变化将为研究质子形状涨落提供一个较好的信号。”摘要：The impact of proton shape fluctuations on the electron-nucleus deeply virtual Compton scattering (DVCS) differential cross-section is studied at the energies of the EIC and LHeC in the framework of the Color Glass Condensate. Proton shape fluctuations are included using the hot spot model, which modifies the dipole scattering amplitude, and a proton shape fluctuation-improved DVCS differential cross-section is obtained. We numerically calculated the DVCS differential cross-sections, and compared the results with those without proton shape fluctuations. We found that the proton shape fluctuations have a certain impact on the electron-proton DVCS differential cross-section in the small momentum transfer region, which leads to the suppression of the cross-section. For the electron-nucleus DVCS process, we found that the proton shape fluctuations influenced on the differential cross-section in the relatively larger momentum transfer region, which renders the peaks and dips of the diffraction to move toward the right. The peaks and dips of the diffraction can be measured with high precision, indicating that this study may provide a suitable method for investigating proton shape fluctuations.关键词：proton structure;color glass condensate;deeply virtual Compton scattering120|132|0更新时间：2025-05-09
- “在高密度磁存储器件领域，研究者基于OOMMF软件模拟了坡莫合金磁纳米线中涡旋与反涡旋畴壁的动力学行为，为磁存储器件设计提供指导。”摘要：The dynamic behavior of vortex and anti-vortex domain walls in Permalloy magnetic nanowires driven by magnetic fields higher than the Walker field was simulated utilizing the OOMMF micromagnetic simulation software. The effects of the magnetic field on the oscillating modes and frequencies of the domain walls for a magnetic nanowire of a fixed size are analyzed. The findings show that, when subjected to a magnetic field stronger than the Walker field, the anti-vortex domain walls will oscillate periodically, and the oscillation modes vary as the magnetic field changes. We also discuss how the oscillation mode and frequency of the domain walls are affected by the size of the magnetic nanowire. The findings indicate that as the size of the magnetic nanowire varies, the domain walls' oscillation mode clearly changes as well. Anti-vortex oscillations are more likely to occur in thin magnetic nanowires, whereas vortex oscillations are the primary oscillations in thick magnetic nanowires. However, the oscillation frequency of the anti-vortex domain wall is more widely controlled, and it can reach higher oscillation frequencies than the vortex. The findings of this study have practical significance for the construction of magnetic logic devices, high-density magnetic memories, and nano-oscillators.关键词：micromagnetic simulation;magnetodynamics;magnetic anti-vortex domain walls;magnetic memory device296|486|0更新时间：2025-05-09
- “在自旋电子学领域，研究者采用脉冲激光沉积法制备含氧空位的WO3薄膜，发现其具有显著的逆自旋霍尔效应和自旋泵浦效应，为自旋电子器件应用提供新思路。”摘要：We prepared WO3 thin films with different oxygen vacancies on (100)-oriented GaAs substrates by pulsed laser deposition. Under the conditions of deposition temperature of 375 °C, oxygen pressure of a 10-30 Pa, and annealing time of 0.5 h, the oxygen-vacancy WO3 thin films with controllable conductivity were prepared, and their resistivity was between that of WO2 and WO3, which increased with increasing of oxygen pressure. In the spin-pumping experiment of the WO3/Py samples, we detected a significant inverse spin Hall voltage and an enhancement of the Gilbert damping factor (up to 3.48×10-3), which indicated that there were inverse spin Hall and spin-pumping effects in the oxygen-vacancy WO3/Py system. Theoretical analysis shows that the spin Hall angle of the oxygen-vacancy WO3 film is as high as -2.934, which is much larger than that of common heavy metal materials. First-principles calculations show that the high spin-to-charge conversion efficiency of the oxygen-vacancy WO3 originates from the occupation of the strong spin-orbit coupling W atom at the Fermi surface. The results show that oxygen-vacancy WO3 thin films have particular potential for application in spintronic devices.关键词：WO3 thin film;oxygen vacancies;ferromagnetic resonance;spin pumping effect;spin Hall angle;first-principles calculations118|116|0更新时间：2025-05-09
- “在机场任务指派领域，专家构建了多目标整数规划模型，设计改进算法，有效解决人员不足问题，为企业运营提供科学依据。”摘要：With the difference between the off-peak season and the occurrence of temporary emergencies, the airport will have a situation of severe shortage of personnel in a certain period when the task volume increases sharply. The studies in this paper do not entirely cover the problem of airport task assignment. The first objective function is to maximize the benefits generated by the task, and the second objective function is to minimize the sum of the differences in the level of qualifications and skills. A multi-objective integer programming model was constructed, and an improved multi-objective memetic algorithm was designed. In the process of solving the problem, the actual data were tested, and the numerical results showed that the accuracy of the built model and improved algorithm was verified compared with CPLEX optimization software. For large-scale examples, when the first objective function value is approximately optimal, the second objective function value is better than the CPLEX solution, with an average optimization of 5.89%. The sensitivity analysis of the coverage rate, shift working hours, and other parameters showed that the setting of different parameters significantly impacted the objective function. This study can not only effectively solve the problem of airport task assignment, but also provide a scientific basis for the actual operational decisions of enterprises.关键词：incomplete coverage;integer programming model;improved multi-objective memetic algorithm83|143|0更新时间：2025-05-09
- “在自动定理证明领域，专家提出了一种子句评估方法，有效提升定理证明效率，为该领域研究提供新思路。”摘要：To solve the problem of effectively selecting clauses in multi-clause deduction, a measurement method for clause influence degree is proposed by analyzing the changes in unification ability before and after deduction. By analyzing the degree of influence of the clause, the number of remaining literals, and the literal deduction ability in the process of multi-clause dynamic deduction, a clause evaluation method based on the comprehensive weight of the clause is proposed, which can effectively control the number of literals in the contradiction separation clause. A multi-clause dynamic deduction algorithm is proposed to effectively select clauses based on this clause evaluation method.The proposed multi-clause dynamic deduction algorithm was applied to the international top first-order logic automated theorem prover Eprover3.1, taking the latest international automated theorem prover competition problems (FOF division) as the test objects. Eprover3.1 with the proposed multi-clause dynamic deduction algorithm outperformed the original Eprover3.1, which solved 18 more theorems than the original Eprover3.1. Regarding solving complex problems, Eprover3.1 with the proposed multi-clause dynamic deduction algorithm can solve 8 theorems that all other provers cannot solve.关键词：multi-clause deduction;clause evaluation;contradiction separation clause;first-order logic;automated theorem prover109|160|0更新时间：2025-05-09
- Tel：027-68756952 Email：whdz@whu.edu.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
- Total Visits：1050059 Visits Today：1567
Bulk Citation
0