Text-Video Retrieval (TVR) is a foundational cross-modal task that aims to align natural language queries with relevant video content, playing a critical role in multimedia search and video understanding. Current methods typically address the cross-modal semantic gap by either augmenting text with generated captions or enhancing video representations at the feature level. However, these strategies often introduce new challenges: generated captions may contain hallucinations that mislead alignment, while dense video representations can retain substantial information redundancy and semantic bias, collectively hindering precise and efficient retrieval. To overcome these limitations, we propose SCAN (Self-Calibrated Textual Anchoring for Dual-Granularity Video Screening). Our framework introduces a dual-granularity video screening mechanism that progressively filters out redundant visual information to obtain semantically condensed video representations. Simultaneously, it employs a self-calibrated textual anchoring module, which leverages the video content to verify and reinforce reliable textual cues, thereby strengthening core text semantics and mitigating caption hallucinations. These two processes are jointly optimized, which aligns text and video semantics in a synergistic manner. Extensive experiments on three benchmarks (MSR-VTT, MSVD, and DiDeMo) demonstrate that SCAN consistently outperforms state-of-the-art methods, validating its effectiveness in reducing hallucinations, eliminating redundancy, and bridging semantic asymmetry for robust text-video retrieval.

The preponderance of traffic accidents takes place within urban intersection areas. It is crucial to extrapolate accident risk maps specifically for these locations to proactively mitigate and prevent future occurrences of traffic accidents. Nevertheless, inferring fine-grained intersection risk remains a challenging task, primarily due to the intricate structure of the road network, variability in scene information, and the stringent requirements for high-quality data. In this work, we propose an end-to-end Adaptive Risk-Feature Aware Fusion Network (ARFAF-Net) based on multimodal data to achieve fine-grained inference of traffic risk maps in intersection areas. Specifically, we introduce a Heterogenous Feature Adaptive Fusion Module to extract complementary features of risk from satellite imagery and streetscape data. The Dynamic Correlation Analysis Module is used to capture large-scale changes in risk in the region to improve multi-scale information perception. In addition, the Macro-Aware Guided Fusion Module is used to introduce macro-satellite image data features to enhance the accuracy of perceiving risks at the periphery of intersections. Ultimately, pixel-level extrapolation maps of intersection crash risk are generated, thereby offering more cost-effective and rational guidance for crash prevention. Both quantitative evaluation and qualitative analysis on real-world datasets demonstrate that the proposed ARFAF-Net achieves superior performance. Finally, the codes and models used in this study for intersection crash risk map inference are available athttps://github.com/gwt-ZJU/ARFAF-Net.

Vision Graph Neural Network (ViG) is designed to recognize different objects through graph-level processing. However, ViG constructs graphs with appearance-level neighbors and neglects the category semantic. The oversight results in the unintentional connection of patches that belong to different objects, thus affecting the distinctiveness of categories in multi-label medical image learning. Since the pixel-level annotations for images are not easily available, category-aware graphs can not be directly built. To solve this problem, we consider localizing category-specific regions using Class Activation Maps (CAMs), an effective way to highlight regions belonging to each category without requiring manual annotations. Specifically, we propose a CAM-interacted Vision GNN (CiV-GNN), in which category-aware graphs are formed to perform intra-category graph processing. CIV-GNN includes a Class-activated Patch Division (CAPD) module, which introduces CAMs as guidance for category-aware graph building. Furthermore, we develop a Multi-graph Interactive Processing (MIP) module to model the relations between category-aware graphs, promoting inter-category interaction learning. Experimental results show that CiV-GNN performs well in surgical tool localization and multi-label medical image classification. Specifically, for m2cai16-localization, CiV-GNN exhibits a 1.43% and 7.02% improvement in mAP50 and mAP50-95, respectively, compared to YOLOv8.

Discretionary lane changing is one of the most important behaviors in vehicle dynamics. The analysis of discretionary lane changing can provide support for human-like autonomous driving system and microscopic traffic simulation. The rule based discretionary lane change decision-making model has the advantages of good interpretability and being in line with human intuitiveness, but the performance of the rule based models are relatively poor. The learning-based discretionary lane changing decision-making models perform better in lane changing analysis than that of the rule based models, while they require a large amount of training data and have poor interpretability. This study proposes a framework for discretionary lane changing decision-making based on cumulative prospect theory and decision tree model, and uses genetic algorithm to calibrate the parameters of the framework. The framework proposed in this study has good interpretability for discretionary lane changing behavior similar to rule-based models, and it can achieve good lane changing prediction performance. The Next Generation Simulation (NGSIM) dataset is used to validate the framework proposed in this study. The results show that the framework proposed in this study can achieve better performance of discretionary lane change prediction than the analysis model based on decision tree and the analysis model based on cumulative prospect theory. When predicting discretionary lane changing behavior 2–0 seconds in advance, the accuracy of the framework in the test set can reach 85.6%.

Merge trees are fundamental topological descriptors with broad theoretical and practical significance, yet their computation, particularly for augmented merge trees, incurs substantial computational and memory overhead. We present BiGMT, a highly scalable parallel algorithm for merge tree construction based on regions of the Morse-Smale segmentation. By exploiting boundary vertex characteristics and the dynamic merging of these regions, BiGMT preserves the essential topology of d-dimensional manifolds and represents the structure as a weighted bipartite graph. This formulation decomposes merge tree construction into fine-grained parallel tasks, significantly reducing traversal cost and memory consumption. Furthermore, BiGMT efficiently constructs augmented merge trees using reverse binary lifting search and parallel connection mechanisms, substantially improving augmentation performance. Extensive experiments on medium- to large-scale scalar field datasets demonstrate substantial GPU acceleration. BiGMT consistently outperforms ExTreeM, PPP, and FTM-Tree on RTX4090 and H200, achieving speedups of up to 23.95× and demonstrating superior parallel efficiency and scalability.

The burgeoning field of video-text retrieval has witnessed significant advancements with the advent of deep learning. However, understanding and matching textual descriptions and video data remains a formidable challenge due to the large information gap across textual and video modalities. As observed, the caption of a video is commonly under-described, lacking expressions of minor characters or local details. Some recent advances have attempted to leverage multimodal Large Language Model (mLLM) to bridge the comprehension gap. However, mLLMs’ potential in enhancing video-text retrieval (VTR) is understudied. This paper aims to fill this research vacancy, analyzing the practical significance and model preferences for utilizing mLLMs in VTR enhancement, as well as investigating the effective integration of mLLM-derived information into the retrieval learning. Based on our analytical insights, we innovatively propose treating mLLM as caption supplements rather than substitutes to bridge the expression gap across modalities. To achieve better cross-modal alignment, we systematically generate diverse variations of videos to construct an elastic visual space. By treating mLLM-supplemented captions as out-of-space points, cross-modal representation learning is accomplished through the optimization of a conical-like representation space. Our model achieves state-of-the-art results on various benchmarks, including MSR-VTT, MSVD, and DiDeMo, and analytical experiments suggest appropriate prompt proposals and indicate our method’s robustness to different mLLMs.

Text-to-video retrieval (T2VR) aims to identify the most semantically relevant video based on a text query. Text queries typically involve diverse visual elements and events in video, making it non-trivial to learn a robust video feature representation for different queries. An abundance of spatial information make the model overwhelmed by redundancy and noisy and struggle to focus on linchpin visual elements. Additionally, without effective guidance, models grapple with connecting temporal information across different frames. In this paper, we introduce a Spatial-Temporal Pooling (STP) method to cohesively capture and unify the inherent spatio-temporal context within videos. For spatial information, STP leverages spatial tags such as entities, scenes, and text as attention prompts, steering the model toward salient visual elements while mitigating the impact of redundancies and noise. For temporal information, STP adopts video narratives summarized in captions as temporal prompts to enhance the model’s perception of events. Experimental results show that our approach has achieved improvements on the MSRVTT(1.9%), MSVD(1.2%), and VATEX(0.7%) datasets compared to the SOTA methods.

ABSTRACT In the era of the Internet of Things and big data, the training of machine learning models has become increasingly demanding due to the vast amounts of data involved. Reducing training time and improving classification accuracy are essential. This article proposes a high‐performance attack detection model (AE‐XGBoost) based on the distributed data parallel processing framework‐Ray. First, a solution called Dynamic Resource Adjustment for Model Training enhances training speed by dynamically adjusting resources, preventing resource idleness or overload, and ensuring optimal resource utilization at each stage. Second, the Dual‐Link Loss Autoencoder algorithm is employed for feature mining, improving anomaly detection and enabling clear visualization of normal and anomalous data. Finally, the data parallel XGBoost method is applied for attack classification. Experimental results on five public large‐scale datasets demonstrate that the proposed model outperforms several well‐established benchmark classification models in both performance and accuracy.

With the rapid development of information technology, cybersecurity issues are becoming increasingly prominent. Existing intrusion detection methods are mainly based on centralized machine learning algorithms, which overlook the data privacy issues of edge devices on the Internet of Things. Therefore, federated learning has been proposed by researchers to address the problems of data privacy and leakage in intrusion detection. However, existing intrusion detection algorithms based on federated learning face the following problems: (1) Lack of a standard network traffic feature set; (2) Clients exhibit data heterogeneity, which severely impacts the training of federated learning; (3) Knowledge distillation-based federated learning requires the server to possess a proxy dataset, which may be impractical in certain situations. To address these issues, this paper proposes an intrusion detection model based on improved generative federated distillation learning (FedGen+). Specifically, we use the original network traffic format as the input to the clients. The server learns a generator to integrate the clients’ label information, then the generator is deployed to clients to generate augmented sample feature representations to train local model. Experimental results demonstrate that the FedGen+ outperforms existing federated learning-based intrusion detection algorithms on the IoT benchmark datasets.

This paper employs lateral and longitudinal trajectory planning to generate candidate trajectories and discards those that do not satisfy the constraints imposed by single-vehicle conditions. Next, a collaborative trajectory combination set for multiple vehicles is derived from the candidate trajectories, with multi-vehicle constraints applied to eliminate combinations that fail to meet the required conditions. The objective function for each candidate trajectory set is first calculated using a single-vehicle objective function, after which a multi-vehicle objective function based on the Kaldor-Hicks improvement principle is constructed. Finally, the paper introduces an improved particle swarm optimization method for multi-vehicle collaborative trajectory planning. The results demonstrate that the dynamic spatiotemporal occupancy growth rate, under varying planning times and frequencies, is at least 19%. Furthermore, the proposed algorithm ensures efficient allocation of travel resources, preventing competition among vehicles that could compromise system feasibility. When verified with HighD trajectory data, the algorithm not only delivers superior optimization results but also exhibits lower standard deviations in dynamic spatiotemporal occupancy and speed compared to real-world data. Finally, the algorithm’s superiority in real-time decision-making and stability is confirmed. Note to Practitioners—In the context of mixed traffic comprising both autonomous and human-driven vehicles, this paper tackles the challenge of coordinating autonomous vehicles to improve traffic efficiency and safety in real-time environments. It presents a promising approach to enhancing cooperative behavior in complex scenarios by integrating real-time data streams to optimize adaptability and ensure equitable driving efficiency across different vehicle types.

Urban public transportation is a complex and open system integral to urban mobility. Its operation is often disrupted by various random factors, necessitating robust scheduling solutions. This study develops a bus robust scheduling model based on mixed-integer linear programming to enhance system resilience. First, an arterial signal coordination model is proposed for mixed traffic environments, enabling autonomous public transport vehicles to traverse intersections without stopping. Second, a demand-deterministic bus scheduling model is constructed, integrating timetables, trajectories, and origin-destination transfer schemes to balance passenger waiting time fairness and efficiency. Third, to address stochastic passenger demand during actual operations, a robust bus scheduling model is developed by incorporating robust constraints. Numerical experiments demonstrate that the demand-deterministic model generates optimal scheduling schemes when passenger demand remains within bus capacity. However, when passenger demand exceeds capacity, the demand-deterministic model becomes infeasible. In such scenarios, the robust scheduling model produces feasible schemes, albeit with reduced optimization, and its robustness can be tuned by adjusting model parameters. Additionally, practical management insights are provided for real-world applications.

Unpaired image-to-image translation aims to translate an image from the source domain to the target domain without paired training data. Some recent works have applied the self-attention to this task and achieved impressive results. Nevertheless, it may generate unsatisfactory results for scenes with complex image content or strong geometric variations between image domains. In this paper, We propose a novel method based on hybrid excitation and adaptive style recombination for unpaired image-to-image translation, which has these main advantages. 1) A hybrid perceptual excitation module is used to capture different ranges of contextual information in complex scenes dynamically. 2) An adaptive cross-attention code recombination module is designed to recombine the content code and style code adaptively according to different positions. 3) We also propose multi-source NCE loss to constrain the generated image content by two aspects: image-level and patch-level. Experiments show that our method achieves more reasonable results than the state-of-the-art methods on several benchmark datasets.

Analyzing discretionary lane-changing (DLC) behavior can provide support for intelligent connected vehicle driving behavior modeling and microscopic traffic simulation. The discrete choice model is the basic model in DLC modeling. However, the discrete choice model cannot fully consider risk preferences during the DLC decision making process. This study harnesses cumulative prospect theory to consider risk preference, constructing a feature vector that includes lane-changing benefits, lane-changing risks, and driving style of drivers. Based on this, a DLC decision making framework based on the discrete choice model is proposed. The framework considers risk preference in the DLC decision making process. It not only retains the interpretability of the discrete choice model but also enhances the predictive performance of DLC. The DLC decision-making framework proposed in this study is validated through the Next Generation Simulation (NGSIM) dataset. The results show that the DLC decision making framework proposed in this study can achieve better performance than the discrete choice model, with an accuracy reaching 77.25% in the test set.

The vigorous development of the Internet of Things (IoT) has generated a massive amount of personal data. How to better protect and use the privacy data of IoT has become the primary problem to be solved today. In this paper, we design a semantic-aware post-quantum searchable encryption scheme to make the use of IoT data more intelligent and secure. Firstly, the topic model and homomorphic encryption technology were used to make the retrieval query expression of ciphertext stronger and the retrieval ability more complete. Secondly, the proxy re-encryption technology was used to make the scheme achieve effective retrieval in the multi-user environment of IoT, so that each user had independent retrieval rights to ensure that there was no possibility of privacy leakage between users. At the same time, to maintain data privacy in all aspects of IoT communication, a simple homomorphic encryption OT protocol was proposed, which provides a secure channel for the user to interact with the topic management server. Finally, the complete scheme construction is based on lattice cryptography, which has proven to be resistant to quantum attacks and has stronger security.

In this study, we introduce a novel hybrid trajectory planning algorithm for autonomous driving, specifically designed to mitigate the risks posed by spoofing attacks on Connected and Autonomous Vehicles (CAVs). The research begins by assessing the safety implications of attacks and developing an adaptive safety model that is grounded in the fundamental assessment of state and decision data. This model incorporates the establishment of a posterior probability distribution for decision data, rooted in the pre-existing prior distribution but adjusted to account for the influence of spoofing attacks. The adjustment is achieved through Bayesian maximum posterior estimation, thereby refining the model to better adapt to potential threats. The adaptive safety model is then optimized dynamically, taking into consideration a set of indices—safety, comfort, and efficiency—that are critical to trajectory planning. In the subsequent phase, we introduce a composite trajectory planning algorithm that integrates a lateral trajectory selection sampling method with a longitudinal trajectory optimization approach. The adaptive safety model is seamlessly integrated into the trajectory planning process, influencing target position selection, the setting of constraints, and the formulation of the optimization objective function. The results demonstrate that the algorithm effectively limits the average standard deviation of lateral displacement to 0.5851 and achieves a significant increase in longitudinal speed growth rate, by up to 7.30%, surpassing the performance of benchmark algorithms. The proposed solution consistently delivers optimal safety and efficiency across various scenarios and under different parameter conditions.

The combination of edge computing and replication technology provides service guarantee for edge applications. However, optimizing replica creation and placement to enhance system performance is challenging due to the limited resources available at the edge. In this context, effective replica management becomes crucial for efficient and reliable edge computing. This study proposes a real-time dynamic replica management model to address the challenges of replica creation and placement in the edge computing environment. Firstly, we design a prediction-based dynamic proactive replica creation algorithm. This algorithm integrates data popularity and node load, utilizing fuzzy membership functions to model data and node states, effectively handling the state uncertainty in certain conditions. It also defines overheating and undercooling similarities to assess the trend of state changes, thereby determining the optimal timing for replica creation. To prevent latency in replica creation, the algorithm employs an Long Short-Term Memory (LSTM) model with a deviation feedback mechanism, which helps prevent lag in replica creation and minimizes unnecessary replica generation. Secondly, we formula replica placement as a multi-objective optimization problem considering the node load and access degree. We use a joint optimization replica placement algorithm that combines Evolutionary Gradient Search (EGS) and Sorting Genetic Algorithm-II to solve the multi-objective replica placement problem. Finally, we conduct extensive experiments on the replica management model. The results demonstrate significant improvements in average response time, effective network utilization rate, storage space utilization rate, and system load balancing, which validate the effectiveness of the proposed method.

Cross-view geo-localization is to retrieve the most relevant images from different views. The biggest challenge is the visual differences between different views and the location shifts in practical applications. Existing methods usually extract fine-grained features of the retrieval target and match them by semantic alignment. The Transformer-based approach can focus on more contextual information than the CNN-based approach and also learn the geometric correspondence between two viewpoint images directly through the location encoding information. However, the existing methods need to fully utilize the information from different viewpoints, and the model needs to understand the context information sufficiently. To address these issues, we propose an efficient method to fully use image information from cross-views and feature fusion, divided into two branches: Aerial-View Local-Feature Cross-Fusion(ALCF) and Multi-View Global-feature Cross-Fusion(MGCF). By observing the characteristics of the aerial and street views, we perform a targeted fusion of global and local features from different viewpoints. In addition, we introduce a multi-view semantic alignment module, which can solve the problem that more noise information is introduced when the aerial view and street view images are semantically aligned. Experiments show that our proposed method achieves excellent performance in both the drone viewpoint target localization and drone navigation tasks on the University-1652 dataset.

With the widespread use of the Tor network, attackers who control malicious relays pose a serious threat to user privacy. Therefore, identifying malicious relays is crucial for ensuring the security of the Tor network. We propose a malicious relay detection model called hybrid multi-scale CNN-LSTM with Attention model (MSC-L-A) for the Tor network. The MSC layer uses one-dimensional convolutional neural networks with different convolution kernels to capture complex multi-scale local features and fuse them. The LSTM layer leverages memory cells and gate mechanisms to control the transmission of sequence information and extract temporal correlation information. The attention mechanism automatically learns feature importance and strengthens the weights of parameters that have a substantial impact on the results. Finally, the Sigmoid function is used to classify the data. Experimental results demonstrate that our proposed model achieves higher prediction accuracy and more accurate classification of Tor relays compared to other baseline models.

In the 3GPP-based cellular vehicle-to-everything (C-V2X) architecture, the Roadside Units (RSU) plays an important role for the enhancement of Quality of Service (QoS) of the vehicular applications. The placement of RSUs has been studied in the literature. However, existing works assume known road traffic distribution with given task demands, which is a simplification of the complex real world situation. In this work, we investigate the optimum RSU placement for C-V2X with uncertain traffic density and task demands. We formulate this RSUs Placement in C-V2X Network (RPCN) problem to minimize the expected vehicle tasks offloading delay through uncertain programming where vehicles positions and tasks are treated as arbitrary stochastic variables. We propose a learning-based algorithm by integrating Stochastic Simulation (SS), Artificial Neural Network (ANN) and meta-heuristic algorithm to determine the placement from real traffic data. The proposed method is an offline design with high practicability. We conducted intensive real-trace driven simulations to demonstrate the effectiveness of our approach on placing RSUs with lower task offloading delay.

In order to better evaluate the anonymity of the Tor network, this paper evaluates the degree of the Tor network anonymity by calculating the probability of routing links being hijacked when users communicate under different node labels. According to the official routing node algorithm, we discuss the results of link establishment from the perspective of node labeling separately and calculate the hijacked rate of different links based on the established evaluation model. The experimental results based on the real data of nodes provided by the official CollecTor website show that the Tor network can resist the intrusion of 10% of malicious nodes, and the anonymity of the Tor network decreases rapidly when the malicious nodes in the routed nodes exceed 20%. In a collection of nodes with different labels, nodes with exit labels have a greater impact on system anonymity when subjected to the same level of attack.

In social networking platforms, the text published by users is usually the most direct way to express users' interests. This paper studies the method of text classification to mine user's interests for friends recommendation. And by fine-tuning the BERT (Bidirectional Encoder Representation from Transformers) pre-training language model to complete text classification tasks. Aiming at the problem of local information missing in the output of BERT pre-training language model. This paper proposes the KBERT-CNN (K_layers Bidirectional Encoder Representation from Transformers and Text Convolutional Neural Networks) text classification model. The model uses the output of the last four layers Transformers of the BERT pre-training language model as text vector, and combines with TextCNN(Text Convolution Neural Network) to build a text classification model. Then this paper uses the probability distribution of user's texts categories to calculate the interest similarity between users to achieve Top-N friends recommendation. Experimental results show that the F1 of the KBERT-CNN text classification model reaches 92.26%, which is better than other text classification models. The Precision of friends recommendation based on text classification is ahead of other content-based friends recommendation methods.

Recent work has shown that convolutional neural network models, especially end-to-end models, perform significant better over traditional methods on stereo matching. However, these models neglect that the information at coarse and fine scales is processed interactively when dealing with matching problems in human visual mechanisms, which can help improve the performance of the model. To solve this problem, we propose CSCNet based on mixed spatial pyramid module and cross-scale correlation volume. In the mixed spatial pyramid module, we propose a way to extract multi-scale context information by mixing pooling and dilated convolution. The cross-scale correlation volume perform cross-computation to obtain full correlation of different scales and the best scale of matching, which reduce the matching ambiguity by imitating the human visual mechanism, and it also provide more similarity information for the subsequent regularization process. Experiments on the KITTI and Scene Flow datasets show that our model outperforms the previous methods.

As a novel network architecture, softwaredefined networking (SDN) decouples control and data planes, which brings a lot of flexibility to the network. But the separation arises some problems, such as controller placement problem(CPP). Controller placement problem is an important task especially in wide area networks, which would directly affect latency and performance of entire network. In this paper, we propose a method based on modified density peaks clustering(MDPC) algorithm to solve controller placement problem, aiming to minimize the average propagation latency between switches and controllers. The entire network is divided into several sub-networks, each of sub-networks is managed by a controller. We conduct the experiments on the Internet2 OS3E topology to evaluate the performance of our method. The experimental results show that our controller placement strategy can significantly reduce latency. Specifically, the average latency can be reduced by 35% compared to K-means and 10% compared to optimized Kmeans.

Pruning effectively reduces the size of neural networks, which facilitates deployment of neural networks in production environment, especially in embedded systems with limited computing resources.In this paper, we propose a convolutional neural network pruning method based on attention mechanism.We add a attention module to model to generate scaling factors for channels.The scaling factors are considered as channels' importance score, thus filters and convolution kernels corresponding to channels with lower importance score are removed.Our method has the ability to learn importance of channels during training, instead of considering only the direct impact of parameters like existing methods.Moreover, it does not depend on any dedicated libraries, so could be combined with other compression methods for better performance.In experiments, we prune about 90% parameters in VGGNet with 0.67% accuracy drop and prune about 50% parameters in ResNet-56 with 1.02% accuracy drop.

In data-intensive cloud collaboration services with tens of thousands of users and million-level resources, means of providing personalized and trust-aware services quickly and simultaneously is a challenging issue. In this paper, we propose Per-trust, a trust-aware and fast resource matchmaking scheme for personalized QoS guaranteeing in collaboration cloud service. First, an integrated and trust-aware service broking architecture is proposed across the collaborative cloud computing environment; this architecture can provide trust computing and personalized resource matchmaking capacities. Then, a resource clustering method is proposed based on the multidimensional properties of cloud resources; this method can accurately, quickly meet the personalized requirements of users. Finally, an innovative algorithm is proposed for the trust computing of service resources based on real-time and dynamic monitoring of data, thereby quickly and effectively providing trust-aware resource matchmaking. Different from existing methods, which focus only on QoS and trust issues, our approach adds a resource clustering step before QoS and trust evaluation. Three key components are organically combined, namely, service broking architecture, resource clustering, and security and QoS-related trust computing. To the best of our knowledge, this paper is the first to construct an integrated solving scheme for cloud resource matchmaking that can simultaneously satisfy the trustworthiness and personalization required by users. Theoretical and experimental results verify the effectiveness of the proposed scheme.

Micro service has achieved good results in recent years due to the advent of docker container technology. Although micro service architecture can solve the shortcomings of the monolithic architecture, yet it brings more complex problems of virtual machine(VM) scheduling. Many related studies are aimed at reducing network traffic consumption or reducing the amount of traffic between VMs. However, the methods proposed by these studies don't work well in the micro service's scenario because of the micro service's design concept and more complex container relationships. In this paper, a container scheduling strategy based on neighborhood division in micro service (CSBND) is proposed. The algorithm takes system load balancing and system response time into account to optimize system performance. Extensive experiments show that our algorithm can effectively reduce the system load imbalance and improve the overall system performance compared with other related algorithms.

Resource scheduling is a key technology of cloud computing. In order to manage the resources in cloud efficiently, it is necessary to use workload prediction techniques for resource management. There are some workload's prediction algorithms for this problem, but they all have problems with accuracy and computational efficiency. In this paper, a new approach for Workload Prediction based on Gated Recurrent Units (GRUWP) was proposed. The approach uses a more reasonable workload model and more suitable neural network model for workload prediction, and it is capable to learn the temporal patterns and long range dependencies on large sequences of arbitrary length. Extensive experiments show that the approach can accurately predict the workload on the physical machine(PM) compared with other widely used workload prediction algorithms.

Neural networks are both computationally intensive and memory intensive, making them difficult to deploy on embedded systems with large number of weights consume considerable storage and memory bandwidth. To address this limitation, pruning is an effective way to compress neural networks with high accuracy. To address this limitation, we proposed a method for optimization on parametric model. This method contains three steps. First, we train the network like conventional training. Next, we prune the unimportant connections and retrain the network to get the sparse weight matrix. Finally, we use Singularly Valuable Decomposition (SVD) to do further compression on the sparse weight matrix. Our experiments on MNIST dataset show that our method has the ability on reducing the model size by 4 times and the accuracy could still be kept over 90%.

Quality of Service plays an increasingly important role during the procedure of web service selection. However, with the rapid growth in the number of web services, it becomes difficult to solve the service selection problem quickly. In order to improve the time cost and optimality of service selection, we propose a large scale transactional service selection approach based on Skyline and Ant Colony Optimization algorithm (ACO) to realize the near-to-optimal QoS service selection. The main idea is to take the advantage of Skyline to reduce candidate services for transactional service selection. We first use Skyline to trim the redundant service, then utilize the Ant Colony Optimization algorithm to select the service from the candidate services. Finally, this approach is evaluated experimentally based on a standard, real dataset as well as synthetically generated datasets. It reveals encouraging results in terms of the quality of solutions.

Providing high trustworthy service is the most fundamental task for any cloud computing platform. Users are willing to deliver their computing tasks and the most sensitive data to cloud data centers, which is based on the trust relationship established between users and cloud service providers. However, with the development of collaboration cloud computing, how to provider fast response for a large number of users' service requests becomes a challenging problem. In order to quickly provide highly trustworthy services, the service platform must efficiently and quickly reply tens of millions of service requests, and automatically match-make tens of thousands of service resources. In this context, lightweight and fast (high-speed, low-overhead) trust computing schemes become the fundamental demand for implementing a trustworthy and collaborative cloud service. In this paper, we propose an innovative and parallel trust computing scheme based on big data analysis for the trustworthy cloud service environment. First, a distributed and modular perceiving architecture for large-scale virtual machines' service behavior is proposed relying on distributed monitoring agents. Then, an adaptive, lightweight, and parallel trust computing scheme is proposed for big monitored data. To the best of our knowledge, this paper is the first to use a blocked and parallel computing mechanism, the speed of trust calculation is greatly accelerated, which makes this trust computing scheme very suitable for a large-scale cloud computing environment. Performance analysis and experimental results verify feasibility and effectiveness of the proposed scheme.

Multi-cloud collaborative environment consists of multiple data centers, which is a typical processing platform for big data. This paper focuses on the trust computing requirement of multi-cloud collaborative services and develops a Data-driven and Feedback-Enhanced Trust (DFET) computing pattern across multiple data centers with several innovative mechanisms. First, a trust-aware service monitoring architecture is proposed based on distributed soft agents to serve as middleware for multi-cloud trust computing and task scheduling. A data-driven trust computation scheme based on multi-indicator monitoring data is then proposed. The integration of several key service indicators into trust computing makes this scheme suitable for service-oriented cloud applications. More importantly, according to the intrinsic relationship among users, monitors, and service providers, we propose an enhanced and hierarchical feedback mechanism that can effectively reduce networking risk while improving system dependability. Theoretical analysis shows that DFET pattern is highly dependable against garnished and bad-mouthing attacks. We also build a prototype system to verify the feasibility of DFET pattern and the experiments yield meaningful observations that can facilitate the effective utilization of DFET in the large-scale multi-cloud collaborative environment.

Person re-identification is an important technique to search a probe person against a set of gallery persons and metric learning methods have shown their effectiveness in matching person images. In this paper, an Integrated Metric Learning with Adaptive Constraints (IMLAC) method is proposed to promote the performance for person re-identification. In the method, the difference and commonness of an image pair are combined to define a novel integrated metric. Considering the complex variations of pedestrian images, a rule of adaptive pairwise constraints is extended for the integrated metric to further enhance separation and reunion between image pairs. Extensive experiments conducted on three person re-identification datasets including VIPeR, PRID450S and GRID indicate that the proposed method outperforms the state-of-the-art methods.

Summary Lattice has become an attractive cryptographic tool due to its potential resistance to quantum attacks, worst‐case hardness, simple computation kind, and flexibility. The pre‐image sample algorithm is the most fundamental algorithm in lattice‐based cryptography for its comprehensive applications in various primitives. Currently, due to Micciancio and Peikert (MP) sample algorithm is the most efficient pre‐image sample algorithm. However, this algorithm also needs massive computations. On the one hand, it expenses the cube of the lattice dimension multiplications over reals to set matrices as Gaussian parameters. On the other hand, it needs complex discrete convolution computations. First, this paper proposes an efficient pre‐image sample algorithm with outputs obeying irregular Gaussian distribution. Two measures are adopted to prevent the leakage of the geometrical property of trapdoor caused by irregular Gaussian outputs. A variant of MP trapdoor is proposed, and a new trapdoor is randomly assembled from a big enough space in each sample. Although still using a matrix as the Guassian parameter, in the proposed algorithm, the computational cost to set Gaussian parameters is zero. Meanwhile, the computational overhead for every sample is far less than that of MP sample algorithm. Second, to demonstrate the security and efficiency of the proposed sample algorithm, a hierarchical identity‐based signature scheme is put forward. This scheme is proved existentially unforgeable against selective identity adaptively chosen‐message attacks. Furthermore, the theoretical analysis shows that the proposed identity‐based signature is more efficient than the existing schemes. Copyright © 2016 John Wiley & Sons, Ltd.

Usually, most of hashing methods for information retrieval have a two-step procedure, embedding the data into a low-dimensional intermediate space and then quantizing them into binary codes. In the hyperplane-based hashing methods, the distance between the data in the intermediate space can replace the Hamming distance to improve the retrieval accuracy. In this paper, a novel asymmetric distance for the hypersphere-based method is proposed to improve the accuracy of similarity search. By showing that the distance in the intermediate space can approximate the Euclidean distance between the data points, more useful information can be taken to improve the retrieval accuracy. According to the characteristics of the hypersphere-based hashing method, various asymmetric distance models are developed and described. Our experiments with two datasets have demonstrated that the proposed method can improve the retrieval accuracy of the hypersphere-based hashing methods significantly and achieve the state-of-the-art recall performance.

Metric learning is an effective method for person re-identification. It utilizes latent factors to find a suitable space for measuring distances. In general, a small number of factors are not powerful enough to match the pedestrians while a large number of factors cause high computational cost. In this paper, to balance this trade-off, a novel diversity regularized distance metric learning method is proposed. For feature representation, the local discriminative features are extracted from the source image and an adjacency maximal constraint is developed to handle the misaligned issue. Then a diversity regularizer is used to learn a metric by making the latent factors uncorrelated so that a small amount of latent factors can preserve effectiveness in measuring distances while reducing the computational burden. Our experimental results show that the proposed method with a small amount of factors can obtain comparative or even better performance compared to the state-of-art methods.

Abstract Signcryption as a cryptographic primitive can carry out signature and encryption simultaneously at a remarkably reduced cost. Identity‐based cryptography is more convenient than public key infrastructure‐based cryptography in certificate management. As a result, identity‐based signcryption has been studied extensively, and many efficient and provably secure constructions have been proposed. However, most of these schemes are based on intractability assumptions from number theory, and these assumptions have been threatened by the booming quantum computation. Therefore, a recent trend in cryptography is to construct cryptosystems that are based on lattice‐based intractability assumptions because of their plausible features of quantum attack resistance. In this paper, several identity‐based signcryption schemes from lattice hardness assumptions are proposed. In the standard model, these schemes are indistinguishable againstinneradaptively chosen ciphertext attacks (IND‐CCA2) and strongly unforgeable againstinnerchosen message attacks. In our construction, it does not matter whether the original encryption scheme used to construct signcryption is deterministic or probabilistic; the resulted signcryption schemes can reach IND‐CCA2 security. To achieve this, we carefully combine three techniques—the identity‐based encryption from lattice due to Agrawal–Boneh–Boyen (EUROCRYPT 2010), the framework of lattice‐based short signature due to Boyen (Public Key Cryptography 2010), and the Canetti–Halevi–Katz (abbr. CHK) technique, with necessary and tailored optimization—for transforming an (ℓ+ 1)‐level indistinguishable under chosen plaintext attack secure hierarchical identity‐based encryption (HIBE) into anℓlevel IND‐CCA2 secure HIBE scheme. In addition, our security proof also contains a more efficient simulation tool that might have separate interest in cryptographic applications. Copyright © 2015 John Wiley & Sons, Ltd.

Oriented by requirement of trust management in multiple cloud environment, this paper presents T-broker, a trust-aware service brokering scheme for efficient matching cloud services (or resources) to satisfy various user requests. First, a trusted third party-based service brokering architecture is proposed for multiple cloud environment, in which the T-broker acts as a middleware for cloud trust management and service matching. Then, T-broker uses a hybrid and adaptive trust model to compute the overall trust degree of service resources, in which trust is defined as a fusion evaluation result from adaptively combining the direct monitored evidence with the social feedback of the service resources. More importantly, T-broker uses the maximizing deviation method to compute the direct experience based on multiple key trusted attributes of service resources, which can overcome the limitations of traditional trust schemes, in which the trusted attributes are weighted manually or subjectively. Finally, T-broker uses a lightweight feedback mechanism, which can effectively reduce networking risk and improve system efficiency. The experimental results show that, compared with the existing approaches, our T-broker yields very good results in many typical cases, and the proposed system is robust to deal with various numbers of dynamic service behavior from multiple cloud sites.

As an extension of NSMP (neighbor supporting multicast protocol), PatchPSMP not only inherits the advantages of NSMP [Seungjoon Lee et al., 2000] and PatchODMRP [Meejeong Lee et al., 2001], but also has its own characters. Therefore, PatchPSMP has the following characters: (1)its route setting up and evaluation policy is same as that of NSMP; (2)its local recovery method is same as that of PatchODMRP; (3)it defines the unforwarding neighbor of forwarding nodes as pool nodes to collect the route information from its received data packets to reduce its local route recovery scope further. When a forwarding node finds a link failure between it and one of its upstream nodes, it floods out an ADVT packet. A pool node can answer its received ADVT packet when it connects to the wanted sources defined in the ADVT packet. By the simulation, we compare the performance of these protocols. PatchPSMP outperforms the others.