As edge computing continues to expand across diverse application domains, the demand for smart applications has grown in both complexity and heterogeneity. Workflow plays a critical role in enhancing the performance and automation capability of such applications. However, traditional workflow modeling and optimization still rely on manual efforts or semi-automated tools. How to achieve automated and adaptive workflow generation while simultaneously optimizing resource allocation in edge environments remains a significant challenge. To address these issues, we propose LLMFlowGO, an LLMdriven system that integrates automated workflow generation and resource optimization in edge computing. By leveraging the semantic understanding and reasoning capabilities of large language models, LLMFlowGO automatically transforms task descriptions into structured workflows and supports service selection and configuration aligned with application constraints. Furthermore, by incorporating edge resource sensing, the system optimizes resource allocation, workflow binding, and execution management to adapt to heterogeneous environments. Experimental evaluations show that LLMFlowGO outperforms baseline methods by achieving an effective trade-off among multiple user QoS objectives without sacrificing convergence speed. LLM-FlowGO is open sourced at https://anonymous.4open.science/r/LLMFlowGo-EDE8sdafavsv1v2f, and the demo video is available at https://anonymous.4open.science/r/LLMFlowGo-EDE8sdafavsv1v2f/LLMFlowGo%20Demo%20Video.mp4.

Edge Computing (EC) enables deep neural network training on distributed data, yet it raises significant privacy concerns, particularly under regulations enforcing the “right to be forgotten”. Federated Unlearning (FU) offers a solution by allowing targeted data unlearning without the need for retraining. In Service-Oriented Computing (SOC) systems, where services are composed dynamically and data flows across multiple decentralized nodes, deploying FU introduces additional challenges. Specifically, the lack of direct access to raw data within loosely coupled services, along with the high communication cost required for coordination among distributed components, significantly hinders effective unlearning. Therefore, we propose D3FU, an efficient service-compatible framework that leverages data-free knowledge distillation to achieve self-contained FU. This framework employs local unlearning through Projected Gradient Descent (PGD), which may initially degrade model performance. To mitigate the resulting bias, we integrate Model-Agnostic Meta-Learning (MAML) techniques to generate task-relevant pseudo-samples, thereby enabling data-free distillation and correcting the gradient updates of the local unlearned model. This process effectively restores model performance while ensuring accurate unlearning. Our experimental results, including evaluations of backdoor attacks, demonstrate that D3FU achieves unlearning effects comparable to retraining from scratch, with a maximum reduction in communication cost by up to 32 times.

Face authentication, as a widely deployed biometric service, suffers from privacy leakage risks caused by the exposure and persistent storage of facial features. However, existing face anonymization methods mainly focus on imagelevel de-identification and do not fully support stable repeated authentication in a protected identity space. This paper presents EVA, an event-conditioned framework for privacy-preserving face authentication services. EVA treats an event as a servicescoped diversification factor and combines it with a userheld key to bind face-derived representations to an eventspecific virtual identity. During enrollment and verification, EVA performs identity processing, biometric stabilization, event-key binding, and cancelable template generation, followed by cloudside virtual identity issuance and protected-domain matching. Preliminary experiments with off-the-shelf recognizers show that EVA suppresses original identity cues while maintaining withinevent consistency for repeated verification. Event or key rotation changes the issued virtual identity, supporting revocable reissuance and reducing cross-event linkage. We further discuss client-side protection, threshold selection, exposure scenarios, and service-level evaluation as future work.

Last mile delivery by drones is a core component of innovative logistics systems, relying heavily on AI models for essential operations such as path planning and object recognition. However, models trained on region specific data often experience significant performance degradation when deployed in unfamiliar environments due to geographic domain shifts. This limitation impedes the rapid deployment of logistics networks and hinders model adaptation. Federated Learning (FL), as a distributed machine learning paradigm, enables multiple clients with diverse data to collaborate in training a global model. Nevertheless, within Mobile Edge Computing (MEC) environments, FL faces critical challenges, including data bias, high drone mobility, and intermittent communication windows between drones and edge servers. This paper proposes FedMO, a mobility aware FL framework for drone based last mile delivery with edge cloud collaboration. FedMO introduces a novel algorithmic insight by treating the drone's flight path as a unified proxy for both communication reliability and data distribution heterogeneity. The framework implements a synergistic three stage selection policy that jointly optimizes connectivity success, data value, and resource efficiency. This effectively transforms mobility from a disruption risk into a diversity enhancing asset. Experimental results using real world drone video datasets demonstrate that FedMO improves convergence speed by approximately 15% compared to baseline methods with only $\mathbf{3 0 \% - 4 0 \%}$ client selection. With equivalent client participation, FedMO achieves a 25% improvement in convergence speed over the FedAvg algorithm.

Uncrewed delivery utilizes autonomous vehicles, such as drones or uncrewed ground vehicles (UGVs), to transport goods, parcels, or other materials without human intervention which can significantly reduce delivery costs and time. Servitised drones and UGVs provide flexible, on-demand delivery services that optimize resource usage, improve logistics efficiency. Drone-as-A-Service (DaaS), as a typical paradigm of uncrewed delivery, can maximize delivery efficiency while minimizing costs. Most existing DaaS systems only support modeling for a single type of delivery service, overlooking the diversity of available delivery services. In fact, the collaboration between different types of heterogeneous delivery services can not only reduce delivery costs but also enhance user satisfaction. However, heterogeneous services require necessitating effective service placement strategies to optimize the allocation and coordination of service resources, ultimately improving delivery efficiency and cost-effectiveness. Therefore, how to generate a suitable service placement plan with the objective of minimizing costs, has become a significant challenge. In this paper, a multi-access edge computing based heterogeneous delivery service framework (HDaaS) is proposed for effectively managing heterogeneous delivery resources. Building upon this framework, we design a cost-aware delivery service placement strategy (CDS) consisting of two phases: recommendation phase using CDSR-GA to identify optimal service and placement phase employing CDSP-RL to generate and refine placement plans. Experimental results show that the CDS strategy can generate delivery service placement plans that effectively reduce service provider costs under deadline constraints by about 29.84% on average.

Artificial Intelligence of Things (AIoT) systems are widely utilized in various domains such as smart logistics and smart health. In AIoT systems, computational tasks earmarked for offloading by IoT devices often have different task priorities and heterogeneous resource requirements. However, these factors will affect the QoS of the solution and the task response time. Therefore, motivated by a typical smart warehouse system, a Multi-access Edge Computing (MEC) based distributed task offloading framework is proposed. Afterward, a delaysensitive mixed-integer programming model is defined to address the imbalanced utilization of computing resources between edge and cloud servers during the offloading process. Finally, based on the attention mechanism, a Priority-Driven Multi-Agent Deep Reinforcement Learning algorithm (PDMA-DRL) is proposed to fulfill real-time resource requirements while reducing the task response time. The PDMA-DRL algorithm consists of two agents, and each agent corresponds to a sub-problem. Comprehensive experimental results demonstrate that our proposed PDMA-DRL algorithm can outperform representative existing methods in generating delay-sensitive task offloading plans and effectively reducing the task response time.

With the rapid expansion of Artificial Intelligence (AI) services, smart devices generate a large amount of user data at the edge network, which urgently needs to be protected while effectively extracting information. Federated learning (FL) is an important technology for handling dispersed data and strict privacy requirements in this context. However, the security threats caused by model inversion attacks and poisoning attacks can affect the mutual trust between the client and server. Yet, for these two types of attacks, the existing defense mechanisms are contradictory in terms of whether the model parameters are publicly disclosed. In addition, the data distribution of the clients is imbalanced which will increase the bias of model, reducing its practicality. To address this issue, this study proposes a dual defense selfbalanced federated learning (DDSFL) framework, aiming to introduce a novel lightweight defense mechanism during the model parameter aggregation stage, combating these two types of attacks simultaneously by applying differential privacy and adjusting learning rates. In addition, this method also integrates a middleware-based reordering algorithm to enhance the robustness of the framework. Experimental results show that DDSFL effectively improves the ability to resist imbalanced data, forged data, and malicious behavior, significantly enhancing the generalization performance and security of the FL system.

In recent years, drone delivery has drawn significant attention for its promising potential in solving the last-mile delivery problem. As a novel service paradigm, Drone-as-a-Service (DaaS) is emerging as an effective way to address service provisioning problems within complex delivery networks. However, existing DaaS composition frameworks often fail to consider the sharing of delivery services and are not applicable to multi-package drone delivery tasks. Meanwhile, as one of the most critical real-world environmental factors for drones, the impact of dynamic wind conditions is not adequately considered by existing studies. This may lead to Quality of Service (QoS) degradation of delivery services. To address the above issues, in this paper, we propose a wind-aware service provisioning strategy for multi-package drone delivery. Given the advantages of Edge Computing (EC) in handling such dynamic factors due to its low latency and high reliability, we first establish a spatio-temporal DaaS model based on service sharing according to the edge-based drone delivery system. Then, we propose a novel wind-aware drone delivery service provisioning strategy for multi-package delivery to minimize energy consumption of drones. The proposed strategy consists of two phases: service sharing and service composition. In the service sharing phase, the service sharing plan is generated by an improved genetic algorithm. In the service composition phase, the edge server dynamically generates the service composition plan through our proposed policy iteration based DaaS composition method. Experimental results using real delivery network and wind data demonstrate that our strategy is able to reduce delivery energy consumption of drone by about 10.1%.

The computer-aided early diagnosis method of Alzheimer's disease (AD) based on structural magnetic resonance imaging (sMRI) has effectively improved the screening efficiency. However, current sMRI-based auxiliary diagnosis is mostly limited to using global information for diagnosis. There are still challenges in using the degree of structural variation in different brain regions and the correlation between brain regions to assist in disease diagnosis. To address these challenges, we propose the MRCAD model based on the interconnection of features from ROIs in various brain regions. This model consists of three parts. Firstly, the multi-brain-region feature extraction module: Extract 90 ROI brain partitions of the brain through a standard brain template, and use a network with shared parameters to accurately extract lesion features. Secondly, the cross-brainregion feature weighting module: Use the brain-region feature correlation method to adaptively enhance the feature weights with a higher degree of correlation, so that the network can pay attention to the correlation between brain regions and adaptively weight according to the contribution of each brain region. Thirdly, the cross-brain-region attention fusion module: Use the attention mechanism to combine clinical scores with the weighted brain-region features to enhance the accuracy of the network. The experimental results of our proposed MRCAD method on the ADNI database show that the prediction accuracy of the MRCAD model for AD, MCI, and CN reaches 95%. Compared with other 3D structural MRI-based AD prediction models, it has a higher prediction accuracy.

Despite domain generalization (DG) has significantly addressed the performance degradation of pre-trained models caused by domain shifts, it often falls short in real-world deployment. Test-time adaptation (TTA), which adjusts a learned model using unlabeled test data, presents a promising solution. However, most existing TTA methods struggle to deliver strong performance in medical image segmentation, primarily because they overlook the crucial prior knowledge inherent to medical images. To address this challenge, we incorporate morphological information and propose a framework based on multi-graph matching. Specifically, we introduce learnable universe embeddings that integrate morphological priors during multi-source training, along with novel unsupervised test-time paradigms for domain adaptation. This approach guarantees cycle-consistency in multi-matching while enabling the model to more effectively capture the invariant priors of unseen data, significantly mitigating the effects of domain shifts. Extensive experiments demonstrate that our method outperforms other state-of-the-art approaches on two medical image segmentation benchmarks for both multi-source and single-source domain generalization tasks. The source code is available at https://github.com/Yore0/TTDG-MGM.

Effective last-mile delivery is pivotal in smart logistics system. While existing delivery network architectures, such as Drone-as-a-Service (DaaS), are capable of enhancing order delivery effectiveness, they often fall short in provisioning diverse delivery services. Furthermore, DaaS-based last-mile delivery systems face challenges from limited payload capacity and range. In this paper, we propose a UAV-UGV integrated network architecture based on Multi-access Edge Computing (MEC), denoted as DaaS+, encompassing the diverse delivery services from both Unmanned Aerial Vehicle (UAV) and Unmanned Ground Vehicle (UGV) for last-mile delivery. To optimize the effectiveness in the delivery process, the intricate overhead and constraints of heterogeneous delivery services are taken into the full consideration. Specifically, we present an energy-aware service model for the UAV-UGV integrated network that considers the deadline constraints of services. Additionally, we address issues of service unavailability during service provisioning. To identify optimal service provisioning plans, we design a novel Energy-Aware Holistic Service Provisioning Mechanism based on Particle Swarm Optimization (ES-PSO), which minimizes delivery energy consumption while adhering to service deadline constraints. Experimental results substantiate the effectiveness of our proposed solution, demonstrating its ability to generate superior service provisioning plans and significantly reduce total delivery energy consumption.

In recent years, there has been a rapid increase in the integration of Internet of Things (IoT) systems into edge computing. This integration offers several advantages over traditional cloud computing, including lower latency and reduced network traffic. In addition, edge computing facilitates the protection of users’ sensitive data by processing it at the edge before transmitting it to the cloud using techniques such as federated learning (FL) and differential privacy (DP). However, these techniques have limitations, such as the risk of user information being obtained by attackers through the uploaded weights/model parameters in FL and the randomness of DP, which limits data availability. To address these issues, this paper proposes a framework called FedShufde (Federated Learning with a Shuffle Model and Differential Privacy in Edge Computing Environments) to protect user privacy in edge computing-based IoT systems, using an unmanned aerial vehicle (UAV) delivery system as an example. FedShufde uses local differential privacy and the shuffle model to prevent attackers from inferring user privacy from information such as UAV’s location, flight conditions, or delivery address. In addition, the network connection between the UAV and the edge server cannot be obtained by the cloud aggregator, and the shuffle model reduces the communication cost between the edge server and the cloud aggregator. Our experiments on a real-world edge-based smart UAV delivery system using public datasets demonstrate the significant advantages of our proposed framework over baseline strategies.

A clustering ensemble provides an elegant framework to learn a consensus result from multiple prespecified clustering partitions. Though conventional clustering ensemble methods achieve promising performance in various applications, we observe that they may usually be misled by some unreliable instances due to the absence of labels. To tackle this issue, we propose a novel active clustering ensemble method, which selects the uncertain or unreliable data for querying the annotations in the process of the ensemble. To fulfill this idea, we seamlessly integrate the active clustering ensemble method into a self-paced learning framework, leading to a novel self-paced active clustering ensemble (SPACE) method. The proposed SPACE can jointly select unreliable data to label via automatically evaluating their difficulty and applying easy data to ensemble the clusterings. In this way, these two tasks can be boosted by each other, with the aim to achieve better clustering performance. The experimental results on benchmark datasets demonstrate the significant effectiveness of our method. The codes of this article are released in http://Doctor-Nobody.github.io/codes/space.zip.

The widespread use of cloud-based face recognition technology raises privacy concerns, as unauthorized access to face images can expose personal information or be exploited for fraudulent purposes. In response, privacy-preserving face recognition (PPFR) schemes have emerged to hide visual information and thwart unauthorized access. However, the validation methods employed by these schemes often rely on unrealistic assumptions, leaving doubts about their true effectiveness in safeguarding facial privacy. In this paper, we introduce a new approach to privacy validation called Minimum Assumption Privacy Protection Validation (Map(2)V). This is the first exploration of formulating a privacy validation method utilizing deep image priors and zeroth-order gradient estimation, with the potential to serve as a general framework for PPFR evaluation. Building upon Map(2)V, we comprehensively validate the privacy-preserving capability of PPFRs through a combination of human and machine vision. The experiment results and analysis demonstrate the effectiveness and generalizability of the proposed Map(2)V, showcasing its superiority over native privacy validation methods from PPFR works of literature. Additionally, this work exposes privacy vulnerabilities in evaluated state-of-the-art PPFR schemes, laying the foundation for the subsequent effective proposal of countermeasures. The source code is available at https://github.com/Beauty9882/MAP2V.

Semi-supervised video object segmentation (VOS) is a highly challenging task, which relies on the initial frame's mask as a segmentation reference in a video sequence to classify each pixel in subsequent frames. However, the guidance provided by the first frame is limited due to the diverse types of segmentation targets and uncertain appearance changes. Consequently, it is crucial to retain useful information during the segmentation process and employ this information for model iteration optimization, enabling the model to better adapt to rapidly changing segmentation objectives. In this work, we propose a multi-scale adaptive model optimization strategy, which incorporates a contextual relevance enhancement module to enforce object correlation by emphasizing feature similarity across adjacent frames. Additionally, we introduce a keyframe discrimination module to deal with the segmentation challenges in scenarios involving significant target changes. Moreover, we also introduce a multi-scale memory screening module to automatically screen and select global-local optimization features for ensuring the model's generalization performance. Extensive experiments show that the proposed method achieves state-of-the-art performance on DAVIS and large-scale Youtube-VOS 2018/2019 datasets without relying on synthetic training data or first-frame fine-tuning.

Workflows are pervasive in software systems where business processes and scientific methods are implemented as workflow models to achieve automated process execution. However, despite the benefit of no/low-code workflow automation, creating workflow models requires in-depth domain knowledge and nontrivial workflow modeling skills, which becomes a hurdle for the proliferation of workflow applications. Recently, Large language models (LLMs) have been widely applied in software code generation given their outstanding ability to understand complex instructions and generate accurate, context-aware code. Inspired by the success of LLMs in code generation, this paper aims to investigate how to use LLMs to automate workflow model generation. We present LLM4Workflow, an LLM-based automated workflow model generation tool. Using workflow descriptions as the input, LLM4Workflow can automatically embed relevant API knowledge and leverage LLM's powerful contextual learning abilities to generate correct and executable workflow models. Its effectiveness was validated through functional verification and simulation tests on a real-world workflow system. LLM4Workflow is open sourced at https://github.com/ISEC-AHU/LLM4Workflow, and the demo video is provided at https://youtu.be/XRQ0saKkuxY.

Edge-Integrated Unmanned Aerial Vehicles (UAVs) delivery systems have demonstrated the advantage of higher efficiency and lower latency in comparison with traditional intelligent delivery systems. But with its rapid development, a series of security and privacy issues have also emerged. For instance, it is of vital importance to maintain data safety due to UAVs exchanging sensitive data with servers through public channels, attackers can easily gain access to sensitive information by launching attacks including man-in-the-middle and impersonation attacks. Additionally, the requirements of frequent authentications between UAVs and edge servers can result in increased computation overhead, while UAVs are fast-moving and resource-constrained, and excessive computational overhead can degrade the user experience. To address these challenges, this paper proposes a Two-Stage Biometric-Assisted Authentication Framework (TBAF) that enhances security and efficiency. In TBAF, a novel secret sharing method is designed to distribute storage biometric templates with protection, ensuring the secret values which are biometric templates can only be accessed by authorized parties. Additionally, the two-stage authentication protocol reduces computation and communication overhead. Extensive formal and informal security analysis confirms the superior performance of the proposed protocol compared to existing solutions.

With the widespread use of Internet of Things (IoT) technology, an enormous number of end devices that request various kinds of cloud services have been connected to the Internet. Multi-access edge computing (MEC) can reduce the service response time by selecting the required edge computing resources closer to the end device. However, MEC-based smart systems require heterogeneous and diverse services support. Taking unmanned aerial vehicle (UAV) last-mile delivery system as an example, there are two types of services required: delivery and computational services. The edge services in MEC environments are distributed and limited. Inefficient service selection plans will affect the quality of services of such smart systems. Therefore, how to design a suitable service selection strategy is a crucial issue for MEC-based smart systems. To address this issue, we propose a service selection framework and a holistic and hybrid service selection ( (HS2)-S-2 ) strategy for MEC-based UAV last-mile delivery systems in real-world UAV last-mile delivery scenarios. This framework considers three important characteristics of UAV delivery systems: diverse service requirements, service availability, and service mobility. The H(2)S(2 )strategy focuses on selecting the optimal delivery and computational services and provides an integrated approach with a static service selection algorithm and a dynamic service re-selection algorithm. The H(2)S(2 )strategy determines the optimal delivery and computational service selection plans with the lowest UAV energy consumption and shortest service response time. We assess the effectiveness and efficiency of the H(2)S(2 )strategy through ablation studies and comparative analyses with diverse representative strategies. The experimental results show that the H(2)S(2 )strategy improves the effectiveness and efficiency of the UAV delivery system by significantly reducing UAV's energy consumption and service response time.

In the realm of cryptography, the implementation of error correction in biometric data offers many benefits, including secure data storage and key derivation. Deep learning-based decoders have emerged as a catalyst for improved error correction when decoding noisy biometric data. Although these decoders exhibit competence in approximating precise solutions, we expose the potential inadequacy of their security assurances through a minimum entropy analysis. This limitation curtails their applicability in secure biometric contexts, as the inherent complexities of their non-linear neural network architectures pose challenges in modeling the solution distribution precisely. To address this limitation, we introduce U-Sketch, a universal approach for error correction in biometrics, which converts arbitrary input random biometric source distributions into independent and identically distributed (i.i.d.) data while maintaining the pairwise distance of the data post-transformation. This method ensures interpretability within the decoder, facilitating transparent entropy analysis and a substantiated security claim. Moreover, U-Sketch employs Maximum Likelihood Decoding, which provides optimal error tolerance and a precise security guarantee.