Knowledge Tracing (KT) diagnoses students’ concept mas- tery through continuous learning state monitoring in education. Existing methods primarily focus on studying behavioral sequences based on ID or textual information. While existing methods rely on ID-based sequences or shallow textual features, they often fail to capture (1) the hierarchical evolution of cognitive states and (2) individualized prob- lem difficulty perception due to limited semantic modeling. Therefore, this paper proposes a Large Language Model Hyperbolic Aligned Knowledge Tracing(L-HAKT). First, the teacher agent deeply parses question semantics and explicitly constructs hierarchical dependencies of knowledge points; the student agent simulates learning behaviors to generate synthetic data. Then, contrastive learning is performed between synthetic and real data in hyperbolic space to reduce distribution differences in key features such as question difficulty and forgetting patterns. Finally, by optimizing hyperbolic curvature, we explicitly model the tree-like hierarchical structure of knowledge points, precisely characterizing differences in learning curve morphology for knowledge points at different levels. Extensive experiments on four real-world educational datasets validate the effectiveness of our Large Language Model Hyperbolic Aligned Knowledge Tracing (L-HAKT) framework.

Graph learning faces major challenges under noisy and sparse supervision, where corrupted labels mislead representation learning and impair generalization. Prior work proposes robust training strategies such as correction, reweighting, and denoising to reduce the influence of noisy labels. However, most methods still optimize directly on training nodes using their possibly corrupted labels as supervision signals. In this work, we propose a prototype-guided framework that replaces direct label supervision over training nodes with semantic supervision derived from class-level prototypes. Each prototype is formed by aggregating representations of nodes sharing the same observed label and serves as a semantic anchor for guiding the classifier. To address the inherent supervision sparsity introduced by limited prototype instances, we introduce a dual-branch mixup strategy that integrates prototypes with high-confidence nodes through intra- and inter-class interpolation, which enhances supervision coverage and improves representation continuity. We further constrain the spatial variance of these samples to promote intra-class compactness. Theoretically, we demonstrate that the constructed prototypes remain aligned with true class semantics under bounded noise rates. Experiments on node classification tasks confirm the effectiveness of our approach under label noise and limited supervision.

The diffusion model with multiple conditions has received widespread attention in the field of drug design due to its high-quality generation ability. However, the paradigm of directly generating new molecules from conditions used in existing work has not accurately fitted the joint distribution of multiple conditions during the generation process. To address this issue, we propose Graph Diffusion Evolution Model(GDEM) for multi conditional molecule generation. GDEM decomposes the process of molecular generation into a chain-like Markov evolution process, continuously adjusting the molecular structure and gradually approaching the true multi-conditional joint distribution. Meanwhile, in order to effectively train this chain evolution generative model, we also propose a two-stage training approximation method to complete the training of intermediate steps. We validated the effectiveness of GDEM on multiple polymer datasets and small molecule datasets, and the results showed that GDEM has advantages in molecular properties and condition control compared to traditional methods.

Heterogeneous graph neural networks (HGNNs) have proven effective at capturing complex relationships in graphs with diverse node and edge types. However, centralized training in HGNNs raises privacy concerns, as sensitive data can be exposed during model training. This risk is further exacerbated when the data is distributed across multiple clients. Federated learning (FL) offers a potential solution by enabling collaborative training without sharing local data. However, existing FL methods for heterogeneous graphs fail to effectively address challenges such as data imbalance and the handling of private edge types. Moreover, existing methods designed for homogeneous graphs are ineffective at addressing the data sparsity issue in heterogeneous graphs. In this article, we propose FedHeGA, a FL framework for heterogeneous graphs that enhances node classification performance while preserving privacy. We tackle data imbalance by integrating heterogeneous graph reconstruction with differential autoencoders to generate semantically coherent node features, improving feature propagation in sparse or imbalanced data. To preserve privacy, we propose a parameter decomposition mechanism that uploads only edge-type-independent global parameters, protecting sensitive local data. Additionally, we address dataset skewness by employing a contrastive learning strategy to align local and global model parameters, which enhances convergence. Experimental results demonstrate that FedHeGA significantly outperforms existing methods on node classification benchmarks, offering an effective solution for federated heterogeneous graph learning.

With the advancement of vehicle-to-vehicle (V2V) ad hoc networks and wireless communication technologies, mobile edge caching has become a key enabler for enhancing network performance and user experience. However, traditional federated learning-based collaborative caching approaches in vehicular scenarios suffer from inadequate client selection mechanisms and limited prediction accuracy, which result in suboptimal cache hit ratios and increased content transmission latency. To address these challenges, we propose a Digital Twin-based Asynchronous Federated Learning-driven Predictive Edge Caching with Deep Reinforcement Learning (DAPR) framework. DAPR employs an intelligent client selection strategy based on asynchronous federated learning, which leverages mobility prediction and data quality assessment to avoid selecting highly mobile clients or clients with low-quality data, thereby significantly improving model convergence efficiency. In addition, we design a GRU-VAE prediction model that uses a Variational Autoencoder (VAE) to capture latent data distribution features and Gated Recurrent Units (GRUs) to model temporal dependencies, thereby substantially enhancing the accuracy of content request prediction. The predicted content popularities are then fed into a deep reinforcement learning-driven caching decision engine to dynamically optimize edge caching resource allocation. Extensive experiments demonstrate that DAPR achieves superior performance in terms of average reward, cache hit ratio, and transmission latency, thereby effectively improving the overall efficiency of vehicular edge caching systems.

Dataset condensation has significantly improved model training efficiency, but its application on devices with different computing power brings new requirements for different data sizes. For sparse graph data with non-Euclidean structures, repeated condensation of each scale may lead to significant computational costs. Thus, condensing multiple scale graphs simultaneously is the core of achieving efficient training in different on-device scenarios. Existing efficient works for multi-scale graph dataset condensation mainly perform efficient approximate computation in scale order (large-to-small or small-to-large scales). However, these two commonly used paradigms for multi-scale graph dataset condensation have serious ''scaling down degradation'' and ''scaling up collapse" problems of a graph. The main bottleneck of the above paradigms is whether the effective information of the original graph is fully preserved when consenting to the primary sub-scale (the first of multiple scales), which determines the condensation effect and consistency of all scales. In this paper, we proposed a novel GNN-centric Bi-directional Multi-Scale Graph Dataset Condensation (BiMSGC) framework, to explore unifying paradigms by operating on both large-to-small and small-to-large for multi-scale graph condensation. Based on the mutual information theory, we estimate an optimal ''meso-scale'' to obtain the minimum necessary dense graph preserving the maximum utility information of the original graph, and then we achieve stable and consistent ''bi-directional'' condensation learning by optimizing graph eigenbasis matching with information bottleneck on other scales. Encouraging empirical results on several datasets demonstrates the significant superiority of the proposed framework in graph condensation at different scales.

Graph neural networks(GNNs) have been demonstrated to depend on whether the node effective information is sufficiently passing. Discrete curvature (Ricci curvature) is used to study graph connectivity and information propagation efficiency with a geometric perspective, and has been raised in recent years to explore the efficient message-passing structure of GNNs. However, most empirical studies are based on directly observed graph structures or heuristic topological assumptions, and lack in-depth exploration of underlying optimal information transport structures for downstream tasks. We suggest that graph curvature optimization is more in-depth and essential than directly rewiring or learning for graph structure with richer message-passing characterization and better information transport interpretability. From both graph geometry and information theory perspectives, we propose the novel Discrete Curvature Graph Information Bottleneck (CurvGIB) framework to optimize the information transport structure and learn better node representations simultaneously. CurvGIB advances the Variational Information Bottleneck (VIB) principle for Ricci curvature optimization to learn the optimal information transport pattern for specific downstream tasks. The learned Ricci curvature is used to refine the optimal transport structure of the graph, and the node representation is fully and efficiently learned. Moreover, for the computational complexity of Ricci curvature differentiation, we combine Ricci flow and VIB to deduce a curvature optimization approximation to form a tractable IB objective function. Extensive experiments on various datasets demonstrate the superior effectiveness and interpretability of CurvGIB.

Due to the explosive growth of electronic education records with highly sensitive nature, the security and authenticity of records have become an urgent issue to be addressed for data sharing. Although the blockchain-assisted searchable attribute-based encryption scheme provides certain trustworthiness for education records sharing, it still suffers from the risk of single key leakage and heavy computational overheads. In this paper, we propose a secure and efficient distributed sharing scheme with attribute-based searchable encryption (SEDS). On the one hand, we design a distributed public key searchable encryption method, which collaboratively generates distributed keys by decentralized blockchain nodes, effectively reducing the risk of single key leakage. On the other hand, we adapt and extend a fast pairing attribute-based encryption method to reduce the computational burden. We evaluate the performance of SEDS through theoretical analysis and experimental verification. The results show that SEDS can not only resist keyword guessing attacks but also reduce encryption and decryption time by 51% and 52% compared to traditional attribute-based encryption schemes, achieving a more secure and efficient education records sharing.

Graph representation learning has become a mainstream method for fraud detection due to its strong expressive power, which focuses on enhancing node representations through improved neighborhood knowledge capture. However, the focus on local interactions leads to imbalanced transmission of global topological information and increased risk of node-specific information being overwhelmed during aggregation due to the imbalance between fraud and benign nodes. In this paper, we first summarize the impact of topology and class imbalance on downstream tasks in GNN-based fraud detection, as the problem of imbalanced supervisory messages is caused by fraudsters' topological behavior obfuscation and identity feature concealment. Based on statistical validation, we propose a novel dual-view graph representation learning method to mitigate Message imbalance in Fraud Detection (MimbFD). Specifically, we design a topological message reachability module for high-quality node representation learning to penetrate fraudsters' camouflage and alleviate insufficient propagation. Then, we introduce a local confounding debiasing module to adjust node representations, enhancing the stable association between node representations and labels to balance the influence of different classes. Finally, we conducted experiments on three public fraud datasets, and the results demonstrate that MimbFD exhibits outstanding performance in fraud detection.

Graph Neural Network-based methods face privacy leakage risks due to the introduction of topological structures about the targets, which allows attackers to bypass the target's prior knowledge of the sensitive attributes and realize membership inference attacks (MIA) by observing and analyzing the topology distribution. As privacy concerns grow, the assumption of MIA, which presumes that attackers can obtain an auxiliary dataset with the same distribution, is increasingly deviating from reality. In this paper, we categorize the distribution diversity issue in real-world MIA scenarios as an Out-Of-Distribution (OOD) problem, and propose a novel Graph OOD Membership Inference Attack (GOOD-MIA) to achieve cross-domain graph attacks. Specifically, we construct shadow subgraphs with distributions from different domains to model the diversity of real-world data. We then explore the stable node representations that remain unchanged under external influences and consider eliminating redundant information from confounding environments and extracting task-relevant key information to more clearly distinguish between the characteristics of training data and unseen data. This OOD-based design makes cross-domain graph attacks possible. Finally, we perform risk extrapolation to optimize the attack's domain adaptability during attack inference to generalize the attack to other domains. Experimental results demonstrate that GOOD-MIA achieves superior attack performance in datasets designed for multiple domains.

Federated graph classification has emerged as a promising paradigm for privacy-preserving graph learning across distributed clients. However, real-world federated scenarios often suffer from severe data heterogeneity and label noise, which significantly degrade model performance. To address these challenges, we propose FedRog, a robust and personalized federated graph neural network framework that improves generalization under non-IID and noisy label settings. FedRog introduces a parameter-aware selection and fine-tuning mechanism to align global and local representations, and a neighbor embedding consistency constraint to enhance robustness against noisy supervision. Furthermore, a fine-grained, importance-guided global aggregation strategy based on Fisher information is employed to mitigate unreliable updates from low-quality clients. We conduct extensive experiments on 16 graph classification datasets under five heterogeneous data partition settings. Results show that FedRog consistently achieves competitive or superior performance compared to 14 baselines in terms of both accuracy and robustness under clean and noisy conditions.

Graph diffusion models have made significant progress in learning structured graph data and have demonstrated strong potential for predictive tasks. Existing approaches typically embed node, edge, and graph-level features into a unified latent space, modeling prediction tasks including classification and regression as a form of conditional generation. However, due to the non-Euclidean nature of graph data, features of different curvatures are entangled in the same latent space without releasing their geometric potential. To address this issue, we aim to construt an ideal Riemannian diffusion model to capture distinct manifold signatures of complex graph data and learn their distribution. This goal faces two challenges: numerical instability caused by exponential mapping during the encoding proces and manifold deviation during diffusion generation. To address these challenges, we propose **GeoMancer**: a novel Riemannian graph diffusion framework for both generation and prediction tasks. To mitigate numerical instability, we replace exponential mapping with an isometric-invariant Riemannian gyrokernel approach and decouple multi-level features onto their respective task-specific manifolds to learn optimal representations. To address manifold deviation, we introduce a manifold-constrained diffusion method and a self-guided strategy for unconditional generation, ensuring that the generated data remains aligned with the manifold signature. Extensive experiments validate the effectiveness of our approach, demonstrating superior performance across a variety of tasks.

We propose a universal video-level modality-awareness tracking model with online dense temporal token learning (called UM-ODTrack). It is designed to support various tracking tasks, including RGB, RGB+Thermal, RGB+Depth, and RGB+Event, utilizing the same model architecture and parameters. Specifically, our model is designed with three core goals: Video-level Sampling. We expand the model's inputs to a video sequence level, aiming to see a richer video context from an near-global perspective. Video-level Association. Furthermore, we introduce two simple yet effective online dense temporal token association mechanisms to propagate the appearance and motion trajectory information of target via a video stream manner. Modality Scalable. We propose two novel gated perceivers that adaptively learn cross-modal representations via a gated attention mechanism, and subsequently compress them into the same set of model parameters via a one-shot training manner for multi-task inference. This new solution brings the following benefits: (i) The purified token sequences can serve as temporal prompts for the inference in the next video frames, whereby previous information is leveraged to guide future inference. (ii) Unlike multi-modal trackers that require independent training, our one-shot training scheme not only alleviates the training burden, but also improves model representation. Extensive experiments on visible and multi-modal benchmarks show that our UM-ODTrack achieves a new SOTA performance.

In the control of discrete-event systems for current-state opacity enforcement, it is difficult to synthesize a supervisor by supervisory control theory (SCT) without explicit formal models of the systems. This study utilizes the reinforcement learning (RL) method to obtain supervisory policies for opacity enforcement in the case when the automaton model of the system is unavailable. The state space of the environment in the RL is dynamically generated through system simulation. Actions are defined according to the control patterns of the SCT. A reward function is proposed to evaluate whether the secret is exposed or not. Then, a sequence of state-action-reward chains are obtained as system simulation goes on. The frameworks of Q-learning and State-Action-Reward-State-Action (SARSA) algorithms are adopted to implement the proposed approach. The goal of the training is to maximize the total accumulative reward by optimizing the action selection in the learning process. Then, an optimal supervisory policy is obtained when the training process converges. Experiments are performed to illustrate the effectiveness of the proposed approach. The contributions are two aspects. Firstly, a supervisor for opacity enforcement is learned by RL training without an explicit formal model of the system. Secondly, the ability of the proposed method in computing supervisory policies without formal models addresses a significant gap in the literature and offers a new direction for research in opacity enforcement in discrete event systems. Note to Practitioners—Supervisory Control Theory (SCT) supplies an effective way to synthesize supervisors, which traditionally handles tasks with explicit system models for current-state opacity enforcement by restricting behavior of systems. However, formal models of systems are often confidential or otherwise unavailable. This paper presents a method for supervisor synthesis via reinforcement learning in the case of lacking formal models of systems. The proposed method leverages the characteristics of control patterns in SCT and optimizes the action selection in the training process through a reward mechanism that evaluates the secrecy of states. The approach can be applied to model-free RL frameworks such as Q-learning and State-Action-Reward-State-Action (SARSA) algorithms. The training is performed as the system simulation goes on. When the training process converges, the optimal policy can be used to enforce opacity for the system. However, Q-table is used to save the Q-value in both Q-learning and SARSA algorithms. In the worst case, the size of the Q-table grows exponentially with the number of states and controllable events increasing. This can lead to memory exhaustion when the system’s scale is large. To make the approach scalable, we will attempt to use DRL to train control policies for opacity enforcement in the future.

Vision-language tracking is a crucial branch of multi-modal object tracking, aiming to jointly locate an object by utilizing visual information and language descriptions. Typically, existing vision-language trackers employ language and visual encoders to extract features from language descriptions and visual information, respectively. Based on these extracted visual and language features, a cross-modal interaction module is used to extract multi-modal features to locate the targets. However, they ignore the differences between visual and language modalities. Due to the lack of pixel-level position information in language descriptions, the positional information of the multi-modal features is greatly weakened by the cross-modal interaction modules. As a result, the vision-language trackers cannot effectively capture subtle changes in the target's positions. To address this problem, we propose a multi-modal hybrid interaction vision-language tracking method (named MHITrack), in which a multi-modal hybrid interaction decoder is designed to enhance the positional information of multi-modal features. The proposed multi-modal hybrid interaction decoder consists of a visual-language interaction module, a multi-level position interaction module, and a hybrid interaction module. Firstly, the multi-level position interaction module is utilized to capture fine-grained position information of the target from multi-level features. Meanwhile, the visual-language interaction module performs cross-modal interaction between visual and language features to obtain multi-modal features. Furthermore, the hybrid interaction module is employed to integrate the multi-modal features with target position information, enhancing the positional information of the multi-modal features. Finally, the proposed tracker can effectively capture subtle changes in the target's positions. Through extensive experiments on four benchmark datasets, namely TNL2k, LaSOT, OTB-Lang, and LaSOText, we demonstrate that the proposed vision-language tracker achieves promising performance compared to existing state-of-the-art vision-language trackers.

In recent years, blockchain oracle, as the key link between blockchain and real-world data interaction, has greatly expanded the application scope of blockchain. In particular, the emergence of the Multi-Data Source (MDS) oracle has greatly improved the reliability of the oracle in the case of untrustworthy data sources. However, the current MDS oracle scheme requires nodes to obtain data redundantly from multiple data sources to guarantee data reliability, which greatly increases the resource overhead and response time of the system. Therefore, in this paper, we propose a Secure and Efficient Multi-data Source Oracle framework (SEMSO), where nodes only need to access one data source to ensure the reliability of final data. First, we design a new off-chain data aggregation protocol TBLS, to guarantee data source diversity and reliability at low cost. Second, according to the rational man assumption, the data source selection task of nodes is modeled and solved based on the Bayesian game under incomplete information to maximize the node's revenue while improving the success rate of TBLS aggregation and system response speed. Security analysis verifies the reliability of the proposed scheme, and experiments show that under the same environmental assumptions, SEMSO takes into account data diversity while reducing the response time by 23.5%.

How to effectively exploit spatio-temporal information is crucial to capture target appearance changes in visual tracking. However, most deep learning-based trackers mainly focus on designing a complicated appearance model or template updating strategy, while lacking the exploitation of context between consecutive frames and thus entailing the when-and-how-to-update dilemma. To address these issues, we propose a novel explicit visual prompts framework for visual tracking, dubbed EVPTrack. Specifically, we utilize spatio-temporal tokens to propagate information between consecutive frames without focusing on updating templates. As a result, we cannot only alleviate the challenge of when-to-update, but also avoid the hyper-parameters associated with updating strategies. Then, we utilize the spatio-temporal tokens to generate explicit visual prompts that facilitate inference in the current frame. The prompts are fed into a transformer encoder together with the image tokens without additional processing. Consequently, the efficiency of our model is improved by avoiding how-to-update. In addition, we consider multi-scale information as explicit visual prompts, providing multiscale template features to enhance the EVPTrack's ability to handle target scale changes. Extensive experimental results on six benchmarks (i.e., LaSOT, LaSOText, GOT-10k, UAV123, TrackingNet, and TNL2K.) validate that our EVPTrack can achieve competitive performance at a real-time speed by effectively exploiting both spatio-temporal and multi-scale information. Code and models are available at https://github.com/GXNU-ZhongLab/EVPTrack.

Hierarchy is an important and commonly observed topological property in real-world graphs that indicate the relationships between supervisors and subordinates or the organizational behavior of human groups. As hierarchy is introduced as a new inductive bias into the Graph Neural Networks (GNNs) in various tasks, it implies latent topological relations for attackers to improve their inference attack performance, leading to serious privacy leakage issues. In addition, existing privacy-preserving frameworks suffer from reduced protection ability in hierarchical propagation due to the deficiency of adaptive upper-bound estimation of the hierarchical perturbation boundary. It is of great urgency to effectively leverage the hierarchical property of data while satisfying privacy guarantees. To solve the problem, we propose the Poincar\'e Differential Privacy framework, named PoinDP, to protect the hierarchy-aware graph embedding based on hyperbolic geometry. Specifically, PoinDP first learns the hierarchy weights for each entity based on the Poincar\'e model in hyperbolic space. Then, the Personalized Hierarchy-aware Sensitivity is designed to measure the sensitivity of the hierarchical structure and adaptively allocate the privacy protection strength. Besides, Hyperbolic Gaussian Mechanism (HGM) is proposed to extend the Gaussian mechanism in Euclidean space to hyperbolic space to realize random perturbations that satisfy differential privacy under the hyperbolic space metric. Extensive experiment results on five real-world datasets demonstrate the proposed PoinDP’s advantages of effective privacy protection while maintaining good performance on the node classification task.

Online contextual reasoning and association across consecutive video frames are critical to perceive instances in visual tracking. However, most current top-performing trackers persistently lean on sparse temporal relationships between reference and search frames via an offline mode. Consequently, they can only interact independently within each image-pair and establish limited temporal correlations. To alleviate the above problem, we propose a simple, flexible and effective video-level tracking pipeline, named ODTrack, which densely associates the contextual relationships of video frames in an online token propagation manner. ODTrack receives video frames of arbitrary length to capture the spatio-temporal trajectory relationships of an instance, and compresses the discrimination features (localization information) of a target into a token sequence to achieve frame-to-frame association. This new solution brings the following benefits: 1) the purified token sequences can serve as prompts for the inference in the next video frame, whereby past information is leveraged to guide future inference; 2) the complex online update strategies are effectively avoided by the iterative propagation of token sequences, and thus we can achieve more efficient model representation and computation. ODTrack achieves a new SOTA performance on seven benchmarks, while running at real-time speed. Code and models are available at https://github.com/GXNU-ZhongLab/ODTrack.

Federated Learning (FL) is a privacy-preserving machine learning paradigm, enabling decentralized devices to collaboratively train models without sharing local data. Traditional FL approaches, however, rely on averaging parameters across clients with homogeneous models, which limits their applicability in scenarios where clients require heterogeneous models. In this paper, we propose FedISMH, a novel approach to address model heterogeneity in FL. Instead of directly applying knowledge distillation, FedISMH clusters clients based on the structural similarities of client models, where clients’ structural features can be extracted through either labeled or unlabeled dataset This allows the proposed model to identify clients with similar model architectures while preserving privacy. Additionally, FedISMH introduces a dynamic mechanism to manage noise clients by aligning them with the most structurally similar clusters, ensuring that their inclusion promote the performance of the cluster. Experimental results on MNIST and SVHN demonstrate that FedISMH consistently outperforms state-of-the-art methods in both IID and Non-IID settings, offering improved accuracy, robustness, and flexibility in heterogeneous FL environments.

Sequential recommendation is a hot research in recommender systems, which Transformer-based models have achieved state-of-the-art performance. However, existing methods lack consideration of historical-level information, leading to ineffective modeling of user preference. To utilize history-level information and enhance item sequence representation infused with historical information, we design a model based on dual bi-directional Transformer encoders and dual Contrastive Learning named DBT4Rec. We first design a dual bi-directional Transformer encoder to capture the relationship between item-level sequences and history-level sequences, then design a dual Contrastive Learning to enhance item sequence representation integrated with history information. Finally, experiment results on three public benchmark datasets show that our model outperforms state-of-the-art models for sequential recommendation.

Sequential recommendations play a crucial role in modern recommender systems because they capture users’ dynamic interests based on his/her historical interactions. Despite the progress of existing methods in sequential recommendation, they only focus on modeling user interaction sequences but ignore multi-perspective sequences information when modeling user preferences, leading to ineffective modeling user preferences. In other words, these methods lack sufficient semantic information to effectively model the one, who considers multi-perspective factors when purchasing an item. To solve this problem, we propose a multi-perspective information and multi-task contrastive learning framework for sequential recommendation, named DIML. DIML can model user preferences by using multi-perspective information, and multi-task contrastive learning can alleviate data sparsity and enhance data quality. Specifically, we first design the information encoding layer, which enriches item semantic information from multi-perspective sequential information to model user preferences more precisely. Then we design a multi-task contrastive learning module to enhance the data quality of multi-perspective information. Finally, experiments on Taobao, JD and MovieLens datasets show that our model is better than the comparison baselines in terms of evaluation metric NDCG, HR and MRR.

Federated learning (FL) enables distributed clients to collaborate on training while storing their data locally to protect client privacy. However, due to data heterogeneity, including issues related to label distributions skew in heterogeneous scenarios, the resulting global model may not be suitable for all clients. In this work, we introduce a personalized federated learning method called pFedPM, which focuses on addressing this challenge of label distributions skew in heterogeneous scenarios. We replace traditional gradient uploading with feature uploading, and introduce a novel feature fusion scheme to learn personalized local model for clients. Specifically, the server receives feature information from clients, aggregates global features, and sends them back to the clients. Clients achieve personalization by fusing local and global features. Furthermore, we introduce a relation network as an additional decision layer, providing a non-linear learnable classifier to predict labels. Through the novel modeling techniques, our proposed method reduces communication costs and supports heterogeneous client models. Experimental results demonstrate that our approach outperforms recent FL methods on the MNIST, FEMNIST, and CIFAR-10 datasets while requiring less communication.

Diffusion models have made significant contributions to computer vision, sparking a growing interest in the community recently regarding the application of it to graph generation. The existing discrete graph diffusion models exhibit heightened computational complexity and diminished training efficiency. A preferable and natural way is to directly diffuse the graph within the latent space. However, due to the non-Euclidean structure of graphs is not isotropic in the latent space, the existing latent diffusion models effectively make it difficult to capture and preserve the topological information of graphs. To address the above challenges, we propose a novel geometrically latent diffusion framework HypDiff. Specifically, we first establish a geometrically latent space with interpretability measures based on hyperbolic geometry, to define anisotropic latent diffusion processes for graphs. Then, we propose a geometrically latent diffusion process that is constrained by both radial and angular geometric properties, thereby ensuring the preservation of the original topological properties in the generative graphs. Extensive experimental results demonstrate the superior effectiveness of HypDiff for graph generation with various topologies.

Graph Contrastive Learning (GCL) has gained extensive attentions due to its success in label scarcity. GCL methods usually utilizes the graph neural network to learn node representation. However, the graph neural network can only aggregate direct neighbor features in each convolutional layer. The contextual dependencies and higher-order structural information among nodes can’t be captured by the local aggregation. In addition, existing GCL methods don’t consider semantic similarity when constructing positive and negative sample pairs of nodes. To address the problems, we propose a Higher-order Semantic-aware Adaptive Graph Contrastive Learning (HSAGCL) method. HSAGCL first extracts the semantic information of higher-order substructures of nodes, thereby integrating indirect neighbor features into node features. Then, HSAGCL adaptively generates augmented views of graphs through self-attention mechanism. The self-attention scores characterize the importance of correlation between nodes, providing a measure for the selection of positive and negative samples. Finally, HSAGCL learn effective graph-level discriminative representations for graph classification by jointly optimizing the contrastive loss and classification loss. Extensive experiments on 5 benchmark datasets show that the proposed HSAGCL achieves significant performance improvements in graph classification, with an average accuracy improvement of 8.94% over state-of-the-art methods.

Abstract Knowledge graph (KG) is introduced as side information into recommender systems, which can alleviate the sparsity and cold start problems in collaborative filtering. Existing studies mainly focus on modeling users’ historical behavior data and KG-based propagation. However, they have the limitation of ignoring noise information during recommendation. We consider that noise exists in two parts (i.e. KG and user-item interaction data). In this paper, we propose Knowledge-aware Dual-Channel Graph Neural Networks (KDGNN) to improve the recommendation performance by reducing the noise in the recommendation process. Specifically, (1) for the noise in KG, we design a personalized gating mechanism, namely dual-channel balancing mechanism, to block the propagation of redundant information in KG. (2) For the noise in user-item interaction data, we integrate personalized and knowledge-aware signals to capture user preferences fully and use personalized knowledge-aware attention to denoise user-item interaction data. Compared with existing KG-based methods, we aim to propose a knowledge-aware recommendation method from a new perspective of denoising. We perform performance analysis on three real-world datasets, and experiment results demonstrate that KDGNN achieves strongly competitive performance compared with several compelling state-of-the-art baselines.

Currently, most studies on content caching strategies for vehicular networks rely on the Zipf distribution model of content popularity. This approach, however, often fails to accommodate the time-varying nature of content popularity and vehicular traffic. Although some studies have begun considering time-varying content popularity, they overlook the implications of time-varying vehicle traffic, which significantly influences content allocation caching policies and could potentially reduce edge server rental costs for service providers. Addressing this gap, our study takes both time-varying content popularity and vehicle traffic data into account. From the perspective of content service providers, we need to strike a tradeoff in optimizing the hit rate (HR) based on the foundation of utility enhancement. Accordingly, we model the problem as a multiobjective issue that jointly optimizes HR, utility, and replacement cost. By employing a scalar function, we transform this multiobjective problem into a single-objective one, allowing for the modulation of optimization objective priorities via weight tuning. We employ two gated recurrent unit networks to predict both vehicle traffic volume and content popularity for the next moment. We use the predicted area traffic flow volume and content popularity as the system state for our cache allocation decision optimization, which is accomplished using a deep reinforcement learning model. We conducted a series of evaluation experiments using real data sets, and our proposed approach significantly outperforms the baseline scheme.

The Practical Byzantine Fault Tolerance (PBFT) protocol-supported blockchain can provide decentralized security and trust mechanisms for the Internet of Things (IoT). However, the PBFT protocol is not specifically designed for IoT applications. Consequently, adapting PBFT to the dynamic changes of an IoT environment with incomplete information represents a challenge that urgently needs to be addressed. To this end, we introduce DA-PBFT, a PBFT dynamic adaptive framework based on a multi-agent architecture. DAPBFT divides the dynamic adaptive process into two sub-processes: optimality-seeking and optimization decision-making. During the optimality-seeking process, a PBFT optimization model is constructed based on deep reinforcement learning. This model is designed to generate PBFT optimization strategies for consensus nodes. In the optimization decision-making process, a PBFT optimization decision consensus mechanism is constructed based on the Borda count method. This mechanism ensures consistency in PBFT optimization decisions within an environment characterized by incomplete information. Furthermore, we designed a dynamic adaptive incentive mechanism to explore the Nash equilibrium conditions and security aspects of DA-PBFT. The experimental results demonstrate that DA-PBFT is capable of achieving consistency in PBFT optimization decisions within an environment of incomplete information, thereby offering robust and efficient transaction throughput for IoT applications.

An$(H,r,M,N)$combination network is a symmetric relay network that involves a central server equipped with$N$files that communicates with$K$users through$H$cache-less intermediate relays, where each user maintains a local cache of size$M$files and is connected to a distinct subset of$r$relays. In this setting, the well-known uniform scheme is proposed by Zewail and Yener via Minimum Distance Separable (MDS) codes. For practical reasons, this paper studies a more general combination network where each distinct subset of$r$relays is connected to$\Lambda $users, referred to as$(H,r,\Lambda,M,N)$dense-user combination network. Although the Zewail-Yener scheme is also feasible for the considered system, it causes high computational complexity since the use of$(H,r)_{q}$MDS code involves expensive multiplication operations in large finite field. In this paper, we aim to design coded caching schemes not only to minimize the worst-case link-load, but also to be implemented over the minimum operation field, i.e., binary field$\mathbb {F}_{2}$. First, we propose a construction which can transform any coded caching scheme for the shared-link model to the considered dense-user combination network. By applying the transformation approach based on the seminal work proposed by Maddah-Ali and Niesen, we present the MAN-based scheme that operates in binary field. To further reduce the link-load under small memory regions, we propose a hybrid scheme that can extend any caching scheme for the original$(H,r,M,N)$combination network to the considered$(H,r,\Lambda,M,N)$dense-user combination network by an ingenious outer-inner construction. From the theoretical analysis of computation complexity, the proposed schemes can significantly reduce the number of bit operations. From numerical comparisons, the link-loads of proposed schemes are close to or even better than that of Zewail-Yener scheme, while significantly reducing the operation field.

Recently, RGB-T tracking methods have made significant progress, demonstrating remarkable capabilities in addressing the complexities of tracking tasks within demanding environments. However, these methods overlook instability of modal validity in real-world scenarios. This limits the model’s ability to understand the correlation between modalities, thereby hindering the model’s ability to fully leverage the synergistic effects of RGB and TIR. To address this challenge, we propose a novel RGB-T tracking model named MCTrack, from the perspective of leveraging correlation among modalities. First, during the feature extraction stage, we design a novel module based on channel matching modeling to construct bidirectional channel context information flow for two modalities. By leveraging information flow, specific modalities correlation information can be transmitted to two modes, augmenting the correlation between the two modes adaptively. Subsequently, after the feature extraction network, the features of each modality are decoded and transformed to generate more correlated feature representations. During this stage, we extract distinctive and collective features by leveraging the correlation among modalities. Then fusing these features and generated search region features specifically for localization. This aids the model in comprehending the correlation between RGB and TIR under complex scenarios, thereby enhancing its ability to capture and utilize key features. Based on extensive experiments conducted on four popular RGB-T tracking benchmarks, our model demonstrates superior performance, particularly showcasing impressive results on the LasHeR dataset with an achieved Precision of 71.6%.

Transformer has achieved impressive progress in visual tracking due to their capability of global modeling, which enables them to learn low-frequency features(i.e., high-level semantic information). However, it seems to overlook the high-frequency features(i.e., low-level texture and edge information) which are crucial to identify different intra-class object instances in the tracking task. To address this issue, we propose a transformer based tracker via frequency fusion perspective that investigated whether high-frequency and low-frequency features can be effectively combined to achieve robust tracking. Specifically, we design a simple yet effective two-stage fusion strategy and use an appropriate frequency fusion strategy in tracking process of each stage so as to make full use of frequency domain information. In the feature extraction stage, we use wavelet decomposition of high-frequency subbands to solve the performance loss caused by the transformer’s catastrophic forgetting of high-frequency information. In the prediction head stage, we use a variety of wavelet decomposition subbands to model the multi-frequency information. The two-stage fusion strategy makes our model extract more balanced and beneficial multi-frequency information, enabling it to effectively capture target texture information and local edge information while also being sensitive to global information. Extensive experiments on six challenging benchmarks (i.e., LaSOT$_{ext}$, UAV123, TNL2K, LaSOT, TrackingNet, and GOT-10k) demonstrates the superior performance of our tracker.

Most of the existing bounding box-based trackers rely on a classification subnetwork and a regression subnetwork to predict the location and scale of the bounding box. They learn the classification subnetwork by processing each sample individually and applying the suggested classification confidence to produce the final prediction. They typically involve heuristic positive sample configurations, which inevitably introduce mislabelled training samples and therefore deteriorate their tracking performance. Moreover, the parallel prediction of the bounding box position and scale may lead to misalignment of classification and regression. To address these issues,we propose a simple yet effective soft constraint-based tracking framework without positive samples (named SoftCT). SoftCT adaptively senses the target’s pixel position through a soft constraint mechanism, which eliminates potential performance gaps caused by artificially marking the target’s pixel position. In addition, SoftCT computes the state of the bounding box by aggregating such positional information, thereby allowing the tracker to avoid misalignment in classification and regression due to uninformed communication. Specifically, SoftCT directly senses the position of the target pixel and fuses this information into the bounding box prediction, rather than requiring explicit annotation or regression of the target pixel. Extensive experiments on six tracking benchmarks including GOT-10k, TrackingNet, LaSOT, UAV123, LaSOText and TNL2K demonstrate that our tracker achieves state-of-the-art performance, confirming its effectiveness and efficiency.

In this paper, we present a simple, flexible and effective vision-language (VL) tracking pipeline, termed MMTrack, which casts VL tracking as a token generation task. Traditional paradigms address VL tracking task indirectly with sophisticated prior designs, making them over-specialize on the features of specific architectures or mechanisms. In contrast, our proposed framework serializes language description and bounding box into a sequence of discrete tokens. In this new design paradigm, all token queries are required to perceive the desired target and directly predict spatial coordinates of the target in an auto-regressive manner. The design without other prior modules avoids multiple sub-tasks learning and hand-designed loss functions, significantly reducing the complexity of VL tracking modeling and allowing our tracker to use a simple cross-entropy loss as unified optimization objective for VL tracking task. Extensive experiments on TNL2K, LaSOT, LaSOT$_{\mathrm{ext}}$and OTB99-Lang benchmarks show that our approach achieves promising results, compared to other state-of-the-arts.

In the tracking literature, foreground and background information have been extensively investigated to discriminate a target from its surrounding background. However, both foreground and background possess their own spatial-temporal correlation relationship that provide significant information to separate the target from its surrounding background, which has been usually ignored by existing work. To address this issue, we propose a bidirectional transductive network based tracker, which incorporates long-range spatial-temporal and bidirectional constraints. Specifically, our tracker consists of two modules, namely the mask generation module (MGM) and the transduction attention module (TAM). MGM aggregates long-range interdependencies of a target along the history frames for generating accurate target masks. TAM retrieves back to the history frames to find patches similar to the current frame, which are then forwarded along with the target masks generated by MGM. In this manner, each position in the current frame can determine its own identity, whether belonging to either the background or the foreground, hence accurately distinguishing the target from its distractors. We conduct systematically experiments and achieve state-of-the-art performance on several benchmarks, obtaining 69.2% AO on GOT-10k and 82.1% on TrackingNet.

A blockchain provides decentralization and trustlessness features for the Industrial Internet of Things (IIoT), which expands the application scenarios of IIoT. To address the problem that blockchains cannot actively obtain off-chain data, the blockchain oracle is proposed as a bridge between the blockchain and external data. However, the existing oracle schemes make it difficult to solve the problem of low quality of service caused by frequent data changes and heterogeneous devices in IIoT, and the current oracle node selection schemes are difficult to balance security and quality of service. To tackle these problems, this paper proposes a secure and reliable oracle scheme that can obtain high-quality off-chain data. Specifically, we first design an oracle node selection algorithm based on a Verifiable Random Function (VRF) and reputation mechanism to securely select high-quality nodes. Second, we propose a data filtering algorithm based on a sliding window to further improve the consistency of the collected data. We verify the security of the proposed scheme through security analysis. The experimental results show that the proposed scheme can effectively select high-quality nodes, reduce data differences, and improve the quality of service of the oracle. In the oracle network with malicious nodes accounting for 10%, the data accuracy rate is increased by about 4%, and the data variance is reduced by about 45% on average.

Coded distributed computing (CDC) introduced by Li et al. can greatly reduce the communication load for MapReduce computing systems. In the cascaded CDC with$K$workers,$N$input files and$Q$output functions, each input file will be mapped by$r$workers and each output function will be computed by$s$workers such that coding techniques can be applied to create multicast opportunities. The main drawback of most existing CDC schemes is that they require the original data to be split into a large number of input files that grows exponentially with$K$, which would significantly increase the coding complexity and degrade the system performance. In this paper, we first use a classical combinatorial structure$t$-design, for any integer$t\geq 2$, to develop a low-complexity and communication-efficient CDC with$r=s$. Our scheme has much smaller$N$and$Q$than the existing schemes under the same parameters$K$,$r$, and$s$; and achieves smaller communication loads compared with the state-of-the-art schemes when$K$is relatively large. Remarkably, unlike the previous schemes that realize on large operation fields, our scheme operates in one-shot communication on the minimum binary field$\mathbb{F}_2$. With a derived lower bound on the communication load under one-shot linear delivery, we show that the$t$-design scheme is asymptotically optimal. Furthermore, we show that our construction method can incorporate the other combinatorial structures that have a similar property to$t$-design. For instance, we use$t$-GDD to obtain another one-shot asymptotically optimal CDC scheme over$\mathbb{F}_2$that has different parameters from$t$-design. Finally, we show that our construction method can also be used to construct CDC schemes with$r\neq s$that have small file number and output function number.

Extensive evidence shows that Graph Neural Networks (GNNs) are vulnerable to adversarial attacks. Previous work has made great efforts to improve the performance and usability of GNNs, but they remain sensitive and vulnerable to local attacks. Local attacks slightly perturb the target nodes, resulting in misclassification of GNNs on these specific nodes. Graph Universal Defense (GUD) is proposed to generate a defense patch to protect important nodes in a graph from local attacks. However, the research on GUD is still in the preliminary stage. The unique patch does not adaptively protect different nodes, and it has weak robustness against strong local attacks. To address this problem, we propose Personalized Graph Universal Defense (PGUD). It maps nodes into several clusters, each corresponding to a personalized defense patch. We further instantiate a class-specific PGUD. From a perspective of effective perturbation at minimal cost, we theoretically derive the potential attacker nodes manipulated by the adversary, and use patches to mitigate their impact on target nodes. Experimental results show that our approach can be efficiently integrated into any GNN for downstream tasks and provide effective protection for different target nodes in a personalized manner. Our method significantly improves the robustness against local attacks compared to prior work and can be easily scaled to large graphs.

Graph Neural Networks (GNNs) have emerged as promising tools in graph semi-supervised learning. They acquire low-dimensional node embeddings for downstream tasks by aggregating and updating features from neighboring nodes. However, in a semi-supervised setting, the availability of labeled node information is limited, resulting in the underutilization of vast amounts of unlabeled node information. Current research integrates contrastive learning with graph semi-supervised learning to better use this unlabeled node information. Nevertheless, the differences between cross-entropy and self-supervised contrastive loss are often overlooked. To address this issue, we propose a novel relaxed graph contrastive semi-supervised learning method. We analyze the difference between nodes correctly classified under cross-entropy loss and self-supervised contrastive loss. Inspired by this, we design a relaxed contrastive loss that considers the node features optimized under these two loss functions, thereby expanding the positive sample set. Furthermore, to ensure the quality of the positive sample set, we introduce a threshold constraint based on feature similarity and prediction results to select more reliable positive samples. The experimental results indicate that our method exhibits competitive performance across most datasets.

Due to the explosion of online educational resources, the protection and management of educational multimedia data have become more challenging. Blockchain has emerged as a promising technology for copyright management due to its decentralized and traceable nature. However, it still faces problems such as difficulties in copyright maintenance and delays in the consensus process, especially for educational multimedia data with large amounts of data, high demand for privacy, and many participants. This paper proposes a secure and trustworthy copyright protection method based on a redactable blockchain to address these issues. We use a decentralized chameleon hash function for copyright maintenance to enable trusted blockchain editing. This can promptly modify infringing copyrights to complete efficient copyright maintenance. For the delay during consensus, we design a consensus mechanism called Proof of Behavior (PoB) based on the Bayesian network (BN), which can achieve fast and reliable consensus by predicting user behavior and selecting highly trusted nodes to participate in consensus. Finally, We conduct simulation experiments to validate the performance of our proposed approach. Specifically, our approach effectively reduces storage space by 5%-10% comparable to other blockchain solutions and enhances system scalability and security.

In recent years, graph neural networks (GNNs) have become a popular semi-supervised learning method for processing graph-structured data. However, traditional graph neural network models rely heavily on labeled data during learning while neglecting the potential of unlabeled data. To further exploit unlabeled data, pseudo-labeling-based methods select high-confidence pseudo-labeling during training and assign unlabeled nodes to join the training accordingly. Unfortunately, there is a confidence bias in the training process with this approach. On the other hand, self-supervised learning trains GNNs models to learn embeddings by mining the supervised information in the data. Since the training process does not introduce labeling information, the discriminative information of classes is not learned. To this end, for node classification in GNNs we propose a novel method that combines self-supervised and semi-supervised approaches, namely SLCSL, which mainly consists of two modules, random neighbors node consistency and class-centered consistency. Random neighbors node consistency explores rich node representations in unlabeled data by aligning random neighbor nodes. To make full use of labeled information, class-centered consistency establishes relationships between labeled and unlabeled nodes, thus enabling unlabeled nodes to be better aware of class information. Finally, experiments are conducted on various real-world datasets, and the results show that SLCSL outperforms the current state-of-the-art methods.

Federated learning (FL) is a scheme that enables multiple participants to cooperate to train a high-performance machine learning model in a way that data cannot be exported. FL effectively protects the data privacy of all participants and reduces communication costs. However, a key challenge for federated learning is the data heterogeneity across clients. In addition, in real FL applications, the class distribution of data is usually unbalanced. Although many researches have been conducted to solve the problem of data heterogeneity, class imbalance problem usually arises along with the heterogeneity data, resulting in the poor performance of the global model. In this paper, a novel FL method (we call it FedEF) is designed for heterogeneous data and local class imbalance problem via optimize feature extractors and classifiers. FedEF optimizes the local feature extractor representation of individual clients through contrastive learning to maximize the consistency of the feature extractor representation trained by the local client and the central server to handle heterogeneous data. Meanwhile, we modified the cross entropy loss in the model, assigned different loss weights to different classes of data, paid more attention to the class with fewer samples in the training process, and corrected the biased classifier to alleviate the problem of class imbalance, thus can improve the performance of the global model. Experiments show that FedEF is an effective solution to FL model obtained under heterogeneous and local class imbalance.

Knowledge graph (KG) as auxiliary information can solve the cold-start and data sparsity problems of recommender systems.However, most existing KG-based recommendation methods focus on how to effectively encode items with that users have interacted into entities and propagate them explicitly, but neglect the relation-level and context-level modeling of collaborative signals.Therefore, it is inevitable to incorporate some unrelated entities while utilizing a propagation strategy, which may weaken part of the recommendation performance.To address this problem, we propose a novel method named Collaborative Relation Context Consistency (CRCC).Compared with other KG-based methods, we model the relation-level and context-level of collaborative signals in a fine-grained manner.Specifically, we segment the user's collaborative knowledge graph to learn related entity information separately to enrich the embedding of users.Moreover, CRCC links the consistency score between the items that users and neighbors have interacted with as the fusion basis, and then we consider the inherent popularity of items while incorporating consistent entities to enhance the embedding representation of items.Extensive experiments on three real-world datasets show that CRCC outperforms several compelling baselines in both CTR prediction and top-K recommendation.

Social recommendation utilizing social relations as auxiliary information can alleviate data sparsity issues of collaborative filtering methods. However, existing social recommendation methods usually focus on the shallow relations of user-item interactions and social networks to mine potential user preferences, which fail to provide a fine-grained consideration of the inhomogeneous influence and the changes in user interest or item attractiveness over time in high-order interactions of multiple relation types. To address these issues, we propose a novel method named Dynamic Social Recommendation with High-matching Inhomogeneous Relations (DSRHIR) to reveal the influence of inhomogeneity in dynamic interactions on social recommendation from multiple perspectives. Specifically, we first model the short-term dynamic and long-term static representations of both users with social relations and similar relations, and items with correlations. Then, we combine their dynamic and static representations during the interaction and the information from the high-order neighbors of these three relations to obtain a graph-augmented representation. Finally, the graph-augmented representation is used to modulate high-matching user interest and item attractiveness. Extensive experiments on two real-world datasets demonstrate the effectiveness of DSRHIR.

As a new paradigm of machine learning, green learning has achieved performance comparable to deep learning in vision tasks, knowledge graph learning, and modeling graph structures. Compared with deep learning, green learning has the advantages of a low carbon footprint, lightweight model and logical transparency. In this paper, to enhance the robustness of green learning, we study the performance of the green model when the training dataset labels are noisy. Through theoretical analysis and experimental verification, we found that noisy labels will not only reduce the performance of green learning, but also amplify the risk of privacy leaks of training dataset members. Therefore, we propose a robust green learning to approach the above two threats. Specifically, we design a multi-stage label-consistency sample selection method to filter out noisy labels in the training dataset by taking advantage of the characteristics of unsupervised representation learning in green learning. In order to further improve the stable classification performance of the model, we propose a feature-level class-balanced data augmentation method in the feature decision learning stage to solve the class imbalance problem caused by sample selection. Finally, a large number of experiments on multiple noisy datasets show that the robust green learning method proposed in this paper can not only enhance the generalization of the model under noisy training labels, but also alleviate the risk of member privacy information leakage.

Graph Neural Networks (GNNs) have various applications in real-life scenarios. However, they are vulnerable to adversarial attacks, especially Graph Injection Attacks (GIAs). The flexible and high-risk GIAs pose a significant threat by injecting malicious nodes into the graph. Regrettably, defense strategies to resist GIAs are still scarce. The current defenders lack comprehensive strategies that can effectively resist a wide range of injection techniques. Particularly, when confronted with highly unnoticeable GIAs, the performance of these defenses is significantly diminished. In this paper, we propose a simple and effective GIA defense method named Faithful INjection DefendER (FINDER). FINDER evaluates the impact of the nodes on their neighbors in the context of message aggregation to identify the potential injected nodes in the graph. By blocking identified injected nodes from interacting with the benign nodes, FINDER prevents the further propagation of misinformation and the spread of adversarial influence throughout the graph. Through extensive experiments conducted on three benchmarks, FINDER demonstrates remarkable performance in accurately identifying injected nodes and defending various GIAs.

Abstract Vulnerability reports are essential for improving software security since they record key information on vulnerabilities. In a report, CWE denotes the weakness of the vulnerability and thus helps quickly understand the cause of the vulnerability. Therefore, CWE assignment is useful for categorizing newly discovered vulnerabilities. In this paper, we propose an automatic CWE assignment method with graph neural networks. First, we prepare a dataset that contains 3394 real world vulnerabilities from Linux, OpenSSL, Wireshark and many other software programs. Then, we extract statements with vulnerability syntax features from these vulnerabilities and use program slicing to slice them according to the categories of syntax features. On top of slices, we represent these slices with graphs that characterize the data dependency and control dependency between statements. Finally, we employ the graph neural networks to learn the hidden information from these graphs and leverage the Siamese network to compute the similarity between vulnerability functions, thereby assigning CWE IDs for these vulnerabilities. The experimental results show that the proposed method is effective compared to existing methods.

Variation of scales or aspect ratios has been one of the main challenges for tracking. To overcome this challenge, most existing methods adopt either multi-scale search or anchor-based schemes, which use a predefined search space in a handcrafted way and therefore limit their performance in complicated scenes. To address this problem, recent anchor-free based trackers have been proposed without using prior scale or anchor information. However, an inconsistency problem between classification and regression degrades the tracking performance. To address the above issues, we propose a simple yet effective tracker (named Siamese Box Adaptive Network, SiamBAN) to learn a target-aware scale handling schema in a data-driven manner. Our basic idea is to predict the target boxes in a per-pixel fashion through a fully convolutional network, which is anchor-free. Specifically, SiamBAN divides the tracking problem into classification and regression tasks, which directly predict objectiveness and regress bounding boxes, respectively. A no-prior box design is proposed to avoid tuning hyper-parameters related to candidate boxes, which makes SiamBAN more flexible. SiamBAN further uses a target-aware branch to address the inconsistency problem. Experiments on benchmarks including VOT2018, VOT2019, OTB100, UAV123, LaSOT and TrackingNet show that SiamBAN achieves promising performance and runs at 35 FPS.

This work considers the multi-access caching system proposed by Hachem et al., where each user has access to$L$neighboring caches in a cyclic wrap-around fashion. We first propose a placement strategy called the consecutive cyclic placement, which achieves the maximal local caching gain. Then under the consecutive cyclic placement, we derive an upper bound on the coded caching gain of any PDA, thus obtaining a lower bound on the rate of PDA-based coded caching schemes. Finally, we construct a class of PDAs under the consecutive cyclic placement, leading to a multi-access coded caching scheme with linear subpacketization, which achieves the derived lower bound on the rate for some parameters; while for other parameters, the achieved coded caching gain is only 1 less than the derived upper bound on the coded caching gain. Analytical and numerical comparisons of the proposed scheme with existing schemes are provided to validate the performance.

Robust tracking has a variety of practical applications. Despite many years of progress, it is still a difficult problem due to enormous uncertainties in real-world scenes. To address this issue, we propose a robust anchor-free based tracking model with uncertainty estimation. Within the model, a new data-driven uncertainty estimation strategy is proposed to generate uncertainty-aware features with promising discriminative and descriptive power. Then, a simple yet effective pyramid-wise cross correlation operation is constructed to extract multi-scale semantic features that provide rich correlation information for uncertainty-aware estimation and thus enhances the tracking robustness. Finally, a semantic consistency checking branch is designed to further estimate uncertainty of output results from the classification and regression branches by adaptively generating semantically consistent labels. Experiments on six benchmarks (i.e., OTB100, VOT2018, VOT2020, TrackingNet, GOT-10K and LaSOT) show the competing performance of our tracker with 130 FPS.

Despite that both local and context information are crucial for robust tracking, existing CNN-based and transformer-based methods mainly focus on one of these aspects. Consequently, the former fails to exploit rich global context information due to the limited receptive field, while the latter suffers from the deficiencies in constructing the local relationship among neighboring regions. To address this issue, we propose the SiamPIN tracker, based on our Parallel Interaction Network. It consists of two effective modules, namely Global Aggregation Block (GAB) and Local Process Block (LPB). GAB perceives the global context to capture the long-range spatial dependency through a transformer-based architecture. Meanwhile, LPB performs local information extraction using a CNN model to retain the detailed appearance information of the target. These two modules are connected consecutively to compose a Trans-Conv unit block, which transmits the global context information to the local feature extraction procedure, hence enables the interaction of global-local information flow. Several such blocks are cascaded so that our model can learn to aggregate local and context information interactively. The proposed tracker achieves state-of-the-art performance on six benchmark datasets, while maintaining a real time running speed.

Occlusion is known as one of the most challenging factors in long-term tracking because of its unpredictable shape. Existing works devoted into the design of loss functions, training strategies or model architectures, which are considered to have not directly touched the key point. Alternatively, we came up with a direct and natural idea that is discarding things that covers the target. We propose a novel occluder-aware representation learning framework to develop this idea. First, we design a local occluders detection module (LODM) to localize the occluders, which works on the principle that discriminates the non-noumenal part from a target based on the general knowledge of this category. An extra dataset and a clustering strategy is proposed to support this general knowledge. Second, we devise a feature reconstruction module to guide the occluder-aware representation learning. With the help of above methods, our localizing occluders tracker, called LOTracker, can learn an occluder-free representation and promote the performance that tracks with occlusion scenarios. Extensive experimental results show that our LOTracker achieves a state-of-the-art performance in multiple benchmarks such as LaSOT, VOTLT2018, VOTLT2019, and OxUvALT.

Social networks are considered to be heterogeneous graph neural networks (HGNNs) with deep learning technological advances. HGNNs, compared to homogeneous data, absorb various aspects of information about individuals in the training stage. That means more information has been covered in the learning result, especially sensitive information. However, the privacy-preserving methods on homogeneous graphs only preserve the same type of node attributes or relationships, which cannot effectively work on heterogeneous graphs due to the complexity. To address this issue, we propose a novel heterogeneous graph neural network privacy-preserving method based on a differential privacy mechanism named HeteDP, which provides a double guarantee on graph features and topology. In particular, we first define a new attack scheme to reveal privacy leakage in the heterogeneous graphs. Specifically, we design a two-stage pipeline framework, which includes the privacy-preserving feature encoder and the heterogeneous link reconstructor with gradients perturbation based on differential privacy to tolerate data diversity and against the attack. To better control the noise and promote model performance, we utilize a bi-level optimization pattern to allocate a suitable privacy budget for the above two modules. Our experiments on four public benchmarks show that the HeteDP method is equipped to resist heterogeneous graph privacy leakage with admirable model generalization.

In the world of cyber security, backdoor attacks are widely used. These attacks work by injecting a hidden backdoor into training samples to mislead models into making incorrect judgments for achieving the effect of the attack. However, since the triggers in backdoor attacks are relatively single, defenders can easily detect backdoor triggers of different corrupted samples based on the same behavior. In addition, most current work considers image classification as the object of backdoor attacks, and there is almost no related research on speaker verification. This paper proposes a novel audio steganography-based personalized trigger backdoor attack that embeds hidden trigger techniques into deep neural networks. Specifically, the backdoor speaker verification uses a pre-trained audio steganography network that employs specific triggers for different samples to implicitly write personalized information to all corrupted samples. This personalized method can significantly improve the concealment of the attack and the success rate of the attack. In addition, only the frequency and pitch were modified and the structure of the attacked model was left unaltered, making the attack behavior stealthy. The proposed method provides a new attack direction for speaker verification. Through extensive experiments, we verified the effectiveness of the proposed method.

Successive subspace learning, a novel green, unsupervised and interpretable machine learning paradigm is widely used in image and point cloud data classification, medical diagnosis, and forgery detection. The multi-stage Saab transform and Channel-wise Saab transform extract low-level to high-level data features for successive subspace learning, but the feature extraction process has some privacy security issues that hinder its further development. Li et al. [1] have proposed corresponding privacy-preserving schemes for the multi-stage Saab transform. However, the above methods would be in the multi-stage Channel-wise Saab transformation has the possibility of failure. Therefore, this paper proposes a noise truncation method based on the differential privacy approach to privacy-protect the Channel-wise Saab transform and designs a green neural network with privacy protection and interpretability using the protected multi-stage Channel-wise Saab transform, Layer-wise relevance propagation method and back propagation algorithm, named SFLW-CNN. Finally, extensive experiments show that the SFLW-CNN method proposed in this paper achieves trade-offs in utility, privacy, and interpretability.

Symmetric Searchable Encryption(SSE) is deemed to tackle the privacy issue as well as the operability and confidentiality in data outsourcing. However, most SSE schemes assume that the cloud is honest but curious. This assumption is not always applicable. In this paper, we propose an efficient SSE scheme based on B+-Tree and Counting Bloom Filter (CBF) which supports secure verification, dynamic updating, and multi-user queries. Comparing with the previous state of the arts, we design the new data structure CBF to support dynamic updating and boost verification. we evaluate our scheme through comprehensive experiments. The results are consistent with our analysis and show that our scheme is secure, and more efficient compared with the previous schemes with the same functionalities.The average performance can be improved by about 20% for both the cloud servers and users when the missing rate of the searching keywords is 20%. And the higher the missing rate is, the more the performance can be improved.

In this paper, we propose a scheme that supports statistic query and authorized access control on an Encrypted Electronic Medical Records Databases(EMDB). Different from other schemes, it is based on Differential-Privacy(DP), which can protect the privacy of patients. By deploying an improved Multi-Authority Attribute-Based Encryption(MA-ABE) scheme, all authorities can distribute their search capability to clients under different authorities without additional negotiations. To our best knowledge, there are few studies on statistical queries on encrypted data. In this work, we consider that support differentially-private statistical queries. To improve search efficiency, we leverage the Bloom Filter(BF) to judge whether the keywords queried by users exists. Finally, we use experiments to verify and evaluate the feasibility of our proposed scheme.

Encrypted image retrieval may return incorrect or incomplete results due to threats from malicious cloud servers. Most of the existing solutions focus on the efficiency and accuracy of retrieval, lack of verification of the completeness of search results, to achieve the reliability of search results and the transparency of the search process, we explore characteristic such as the decentralization and tamper-proof of blockchain, proposed a blockchain-based encrypted image retrieval scheme. This scheme stores the encrypted index on the blockchain (Ethereum), through the blockchain consensus mechanism and the function of searching on the smart contract, ensures the integrity and correctness of search results, then outsources the corresponding encrypted images to the cloud server to reduce storage cost, and designs a double-layer index structure using the bag of visual word model and simhash in the process of image similarity index. Experiments show that the reliability, high retrieval efficiency, and precision of the scheme also have a good privacy protection effect.

Despite the great success of Siamese-based trackers, their performance under complicated scenarios is still not satisfying, especially when there are distractors. To this end, we propose a novel Siamese relation network, which introduces two efficient modules, i.e. Relation Detector (RD) and Refinement Module (RM). RD performs in a meta-learning way to obtain a learning ability to filter the distractors from the background while RM aims to effectively integrate the proposed RD into the Siamese framework to generate accurate tracking result. Moreover, to further improve the discriminability and robustness of the tracker, we introduce a contrastive training strategy that attempts not only to learn matching the same target but also to learn how to distinguish the different objects. Therefore, our tracker can achieve accurate tracking results when facing background clutters, fast motion, and occlusion. Experimental results on five popular benchmarks, including VOT2018, VOT2019, OTB100, LaSOT, and UAV123, show that the proposed method is effective and can achieve state-of-the-art results. The code will be available at https://github.com/hqucv/siamrn

Deep learning has been extensively applied in many fields, such as image segmentation, voice recognition, automatic language translation. However, many malicious attackers attempt to attack the model which was trained to accomplish a deep learning assignment via various schemes. Recently, differential privacy technology has been proposed to defend against such attacks via sacrificing the accuracy of model. Therefore, many optimization methods have been proposed to reduce the overall privacy cost, and aim to seek a tradeoff between privacy and utility. In this paper, we propose an approach based on the cluster technology to get a tighter clipping bound for differential privacy deep learning model. In addition, we quantify the clipping bound with an objective function of standard deviation and prove our scheme in an analytically way. A large number of experiments setting on real-datasets demonstrate that our adaptive clipping bound method is better than the previous method which sets the clipping bound constantly.

K-Nets is a deterministic clustering algorithm based on the network structure. It can automatically detect the sym-metric structure in the data and can be used to process clusters of different sizes, shapes or a specific number. However, K-Nets has the following shortcomings: (1) the clustering result is more sensitive to the manually input parameter K, so the accuracy will be affected; (2) the algorithm only considers the average distance of K-nearest neighbors, which may lead to some wrong distribution center points in the dataset with large density difference or the same score values during calculation; (3) it does not consider the privacy leakage during the clustering process. To solve the above problems, we propose a differential privacy protection method in adaptive K-Nets clustering, called ADP-K-Nets. Firstly, for reducing the influence of the parameters on the result, the natural eigenvalues are adaptively obtained through the characteristic of the natural neighbors and used as parameter values to find data points. Then we define a new method for calculating the score, which can solve the problem of incorrectly selecting cluster centers when there are large density differences or conflicts in the calculation process. Also, the Laplace noise is added in calculating the local density of every data point to protect data privacy. Experimental results show that our method ensures the performance of clustering compared with some existing algorithms.

With the vigorous development of artificial intelligence technology, various engineering technology applications have been implemented one after another. The gradient descent method plays an important role in solving various optimization problems, due to its simple structure, good stability, and easy implementation. However, in multinode machine learning system, the gradients usually need to be shared, which will cause privacy leakage, because attackers can infer training data with the gradient information. In this paper, to prevent gradient leakage while keeping the accuracy of the model, we propose the super stochastic gradient descent approach to update parameters by concealing the modulus length of gradient vectors and converting it or them into a unit vector. Furthermore, we analyze the security of super stochastic gradient descent approach and demonstrate that our algorithm can defend against the attacks on the gradient. Experiment results show that our approach is obviously superior to prevalent gradient descent approaches in terms of accuracy, robustness, and adaptability to large-scale batches. Interestingly, our algorithm can also resist model poisoning attacks to a certain extent.

Set-valued data and social network provide opportunities to mine useful, yet potentially security-sensitive, information. While there are mechanisms to anonymize data and protect the privacy separately in set-valued data and in social network, the existing approaches in data privacy do not address the privacy issue which emerge when publishing set-valued data and its correlative social network simultaneously. In this paper, we propose a privacy attack model based on linking the set-valued data and the social network topology information and a novel technique to defend against such attack to protect the individual privacy. To improve data utility and the practicality of our scheme, we use local generalization and partial suppression to make set-valued data satisfy the grouped ρ-uncertainty model and to reduce the impact on the community structure of the social network when anonymizing the social network. Experiments on real-life data sets show that our method outperforms the existing mechanisms in data privacy and, more specifically, that it provides greater data utility while having less impact on the community structure of social networks.

Feedforward-designed convolutional neural network (FF-CNN) is an interpretable network. The parameter training of the model does not require backpropagation (BP) and optimization algorithms (SGD). The entire network is based on the statistical data output by the previous layer, and the parameters of the current layer are obtained through one-pass manner. Since the network complexity under the FF design is lower than the BP algorithm, FF-CNN has better utility than the BP training method in the directions of semi-supervised learning, ensemble learning, and continuous subspace learning. However, the FF-CNN training process or model release will cause the privacy of training data to be leaked. In this paper, we analyze and verify that the attacker can obtain the private information of the original training data after mastering the training parameters of FF-CNN and the partial output responses. Therefore, the privacy protection of training data is imperative. However, due to the particularity of the FF-CNN training method, the existing deep learning privacy protection technology is not applicable. So we proposed an algorithm called differential privacy subspace approximation with adjusted bias (DPSaab) to protect the training data in FF-CNN. According to the different contribution of the model filters to the output response, we design the privacy budget allocation according to the ratio of the eigenvalues and allocate a larger privacy budget to the filter with a large contribution, and vice versa. Extensive experiments on MNIST, Fashion-MNIST, and CIFAR-10 datasets show that DPSaab algorithm has better utility than existing privacy protection technologies.

Many approaches have been proposed for publishing useful information while preserving data privacy. Among them, the privacy models of identity-reserved (k, l)-anonymity and identity-reserved $$(\alpha , \beta )$$ -anonymity have been proposed to handle the situation where an individual could have multiple records. However, the two models fail to prevent attribute disclosure. To this end, we propose two new privacy models: enhanced identity-reserved l-diversity and enhanced identity-reserved $$(\alpha , \beta )$$ -anonymity. Moreover, to implement the two privacy models we design a general anonymization algorithm, called DAnonyIR, with clustering technique by calling different decision functions, which can decrease the information loss caused by generalization. Further, we compare DAnonyIR concerning our two privacy models with existing generalization method GeneIR concerning identity-reserved (k, l)-anonymity and identity-reserved $$(\alpha , \beta )$$ -anonymity, respectively. The experimental results show that our two approaches provide stronger privacy preservation, and their information loss and relative error ratio of query answering are less than those of GeneIR.

Transactional data with attributes of multiple types may be extremely useful to secondary analysis (e.g., learning models and finding patterns). However, anonymization of such data is challenging because it contains multiple types of attributes (e.g., relational and set-valued attributes). Existing privacy-preserving techniques are not applicable to address this problem. In this paper, we propose a novel graph-based multifold model to anonymize data with attributes of multiple types. Under this model, such data are modelled as a graph, and multifold privacy is guaranteed through fuzzing on sensitive attributes and converting associations among items into an uncertain form. Specifically, we define a multi-objective attack model in a graph and devise a safety parameter and algorithm to prevent such attacks. Experiments have been performed on real-life data sets to evaluate the performance.

Data collection by social networking applications offers many opportunities for mining information, which provides a better understanding of social structures and their dynamic structures. Anonymization of social networks before they are published or shared is particularly important, since social network data usually contain much sensitive information on individuals. In this paper, we address the privacy problems of dynamic releases of social networks. We re-define the label-neighborhood attack model in dynamic social network releases. An adversary can use one-hop neighbor's network structure and label as background knowledge to identity the victim to learn more sensitive information. We propose a dynamic-l-diversity anonymized method to resist attacks. Experiments show that the proposed approach can retain much of the characteristics of the network while providing high utility.

Multimodal transaction data mining has received a great deal of attention recently. Protection of private information is an essential requirement of data analysis. Existing work on privacy protection for transaction data usually focus on a single mode dataset. The existing privacy-preserving methods cannot be used directly to address privacy issues for multimodal data integration, since information leakage may be caused by data correlations among multiple heterogeneous datasets. In this work, we address privacy protection on the integration of transaction data and trajectory data. We first demonstrate a privacy leakage model caused by integration of multimodal datasets, where integrated data are modeled as a tree. To address the identity disclosure of trajectories, we partition location sequences to meet privacy demands, and copy locations to offset information loss caused by partition; then, to deal with the sensitive item disclosure of transactions, we use suppression technique to eliminate sensitive association rules. Consequently, we propose a k m -anonymity- ρ -uncertainty privacy model to protect the privacy information in integrating transaction data with trajectory data in a tree-structured data model. Finally, we perform experiments on two synthetic integration datasets, and analyze privacy and information loss under varying parameters.

With the popularity of social networks, privacy issues with regard to publishing social network data have gained intensive focus from academia. We analyzed the current privacy-preserving techniques for publishing social network data and defined a privacy-preserving model with privacy guarantee [Formula: see text]. With our definitions, the existing privacy-preserving methods, [Formula: see text]-anonymity and randomization can be combined together to protect data privacy. We also considered the privacy threat with label information and modify the [Formula: see text]-anonymity technique of tabular data to protect the published data from being attacked by the combination of two types of background knowledge, the structural and label knowledge. We devised a partial [Formula: see text]-anonymity algorithm and implemented it in Python and open source packages. We compared the algorithm with related [Formula: see text]-anonymity and random techniques on three real-world datasets. The experimental results show that the partial [Formula: see text]-anonymity algorithm preserves more data utilities than the [Formula: see text]-anonymity and randomization algorithms.

Robustness and discrimination are two of the most important objectives in image hashing. We incorporate ring partition and invariant vector distance to image hashing algorithm for enhancing rotation robustness and discriminative capability. As ring partition is unrelated to image rotation, the statistical features that are extracted from image rings in perceptually uniform color space, i.e., CIE L*a*b* color space, are rotation invariant and stable. In particular, the Euclidean distance between vectors of these perceptual features is invariant to commonly used digital operations to images (e.g., JPEG compression, gamma correction, and brightness/contrast adjustment), which helps in making image hash compact and discriminative. We conduct experiments to evaluate the efficiency with 250 color images, and demonstrate that the proposed hashing algorithm is robust at commonly used digital operations to images. In addition, with the receiver operating characteristics curve, we illustrate that our hashing is much better than the existing popular hashing algorithms at robustness and discrimination.

Driven by mutual benefits, there is a demand for transactional data sharing among organizations or parties for research or business analysis purpose. It becomes an essential concern to provide privacy-preserving data sharing and meanwhile maintain data utility, due to the fact that transactional data may contain sensitive personal information. Existing privacy-preserving methods, such as k-anonymity and l-diversity, cannot handle high-dimensional sparse data well, since they would bring about much data distortion in the anonymization process. In this paper, we use bipartite graphs with node attributes to model high-dimensional sparse data, and then propose a privacy-preserving approach for sharing transactional data in a new vision, in which the bipartite graph is anonymized into a weighted bipartite graph by clustering node attributes. Our approach can maintain privacy of the associations between entities and resist certain attackers with knowledge of partial items. Experiments have been performed on real-life data sets to measure the information loss and the accuracy of answering aggregate queries. Experimental results show that the approach improves the balance of performance between privacy protection and data utility.

Modern distributed applications are embedding an increasing degree of dynamism, from dynamic supply-chain management, enterprise federations, and virtual collaborations to dynamic resource acquisitions and service interactions across organizations. Such dynamism leads to new challenges in security and dependability. Collaborating services in a system with a Service-Oriented Architecture (SOA) may belong to different security realms but often need to be engaged dynamically at runtime. If their security realms do not have a direct cross-realm authentication relationship, it is technically difficult to enable any secure collaboration between the services. A potential solution to this would be to locate intermediate realms at runtime, which serve as an authentication path between the two separate realms. However, the process of generating an authentication path for two distributed services can be highly complicated. It could involve a large number of extra operations for credential conversion and require a long chain of invocations to intermediate services. In this paper, we address this problem by designing and implementing a new cross-realm authentication protocol for dynamic service interactions, based on the notion of service-oriented multiparty business sessions. Our protocol requires neither credential conversion nor establishment of any authentication path between the participating services in a business session. The correctness of the protocol is formally analyzed and proven, and an empirical study is performed using two production-quality Grid systems, Globus 4 and CROWN. The experimental results indicate that the proposed protocol and its implementation have a sound level of scalability and impose only a limited degree of performance overhead, which is for example comparable with those security-related overheads in Globus 4.

In this paper, we propose a novel approach for composing existing web services to satisfy the correctness constraints to the design, including freeness of deadlock and unspecified reception, and temporal constraints in Computation Tree Logic formula. An automated synthesis algorithm based on learning algorithm is introduced, which guarantees that the composite service is the most general way of coordinating services so that the correctness is ensured. We have implemented a prototype system evaluating the effectiveness and efficiency of our synthesis approach through an experimental study.

Modern distributed applications are embedding an increasing degree of dynamism, from dynamic supply-chain management, enterprise federations, and virtual collaborations to dynamic service interactions across organisations. Such dynamism leads to new security challenges. Collaborating services may belong to different security realms but often have to be engaged dynamically at run time. If their security realms do not have in place a direct cross-realm authentication relationship, it is technically difficult to enable any secure collaboration between the services. A typical solution to this is to locate at run time intermediate realms that serve as an authentication-path between the two separate realms. However, the process of generating an authentication path for two distributed services can be very complex. It could involve a large number of extra operations for credential conversion and require a long chain of invocations to intermediate services. In this paper, we address this problem by presenting a new cross-realm authentication protocol for dynamic service interactions, based on the notion of multi-party business sessions. Our protocol requires neither credential conversion nor establishment of any authentication path between session members. The correctness of the protocol is analysed, and a comprehensive empirical study is performed using two production quality grid systems, Globus 4 and CROWN. The experimental results indicate that our protocol and its implementation have a sound level of scalability and impose only a limited degree of performance overhead, which is comparable with those security-related overheads in Globus 4.

Security in collaborative groups is an active research topic and has been recognized by many organizations in the past few years. In this paper, we propose a fine-grained and attribute-based access control framework for our key project, CROWN grid. To avoid single point of failure and enhance scalability of the system, we employ a distributed delegation authorization mechanism. We successfully implement our proposed access control in CROWN grid, and evaluate this approach by comprehensive experiments.