Equipment manufacturing industry plays a pivotal role in a nation's economy, national development, and technological innovation. Traditional equipment manufacturing enterprises often face the problem of untimely information sharing and inconsistent focus of various departments in the process of operation, which causes difficulties in analysis and conflict in decision-making. Considering the subjectivity of interdepartmental decisions, reinforcement learning is introduced to analyze the decisions of various departments and get the decisions that best meet the needs of the enterprise. In addition, due to the instability of the internal and external environment of the enterprise, the decision obtained only by the machine usually cannot meet the development requirements of the enterprise, so human–machine coordination is used to adapt to the environmental changes. Therefore, a decision conflict resolution for equipment manufacturing enterprises based on human–machine coordination is proposed in this article. The resolution of conflicts in the context of deep human–computer collaborative decision-making is achieved through the incorporation of expert knowledge twice within the framework of reinforcement learning, thereby expressing preferences for both production factors and decision model structures. The experiments indicate that our model achieves superior performance in resolving decision conflicts.

In the field of intelligent manufacturing, the vast and heterogeneous nature of data across different departments poses significant challenges for collaborative decision-making. Direct data sharing often leads to severe privacy and security concerns. Although federated learning offers a promising solution, issues such as participant selection and privacy protection remain inadequately addressed. During model training, it is crucial to minimize the influence of low-quality participants and prevent the inference of sensitive information through gradient analysis, which could threaten model performance or data privacy. To address these challenges, this paper proposes a robust federated learning model, Fed-RDP, based on participant contribution evaluation and personalized differential privacy. The model leverages blockchain technology for secure data flow and storage, implemented through smart contracts. Historical parameters are submitted to evaluate contributions based on the concept of the least core, with real-time updates to reputation scores. When participants upload their local models, the scores are used as weights to aggregate and update the global model. Additionally, personalized differential noise is added to the uploaded gradients based on participant scores, preserving privacy while maximizing the utility of the data. Experimental results demonstrate that this approach effectively identifies low-quality participants, optimizes evaluation time, and protects privacy through personalized differential noise.

The disadvantages of personnel tracking methods are such that traditional manual methods are inefficient, while internet-based tracking methods overload users with information. Manual viewing of surveillance videos is time-consuming and labor-intensive, and real-time search methods based on machine vision require excessive computation. Therefore, this paper proposes a specific pedestrian tracking method based on location prediction and edge computing. This method involves retrieving videos from relevant areas based on the given personnel image information, generating an image set containing the target and the target’s approximate trajectory, and performing personnel tracking. The proposed pedestrian tracking framework treats surveillance videos as three-dimensional spatiotemporal maps, transforming the pedestrian tracking task into a map search problem. The framework consists of two algorithms: a location prediction algorithm based on road network information and a multistep trajectory tracking algorithm for tracking specific pedestrians in an edge computing environment. The location prediction algorithm determines the target’s location and predicts the target’s next position, while the multistep trajectory tracking algorithm enables trajectory tracking and distributed training of the model in the context of an edge network. Experimental results demonstrate that the proposed solutions are highly practical and perform well in personnel tracking compared to previous approaches.

As the development of the Internet of Things (IoT), the security of images in social networks is attracting more and more attention. Among the various encryption methods, thumbnail-preserving encryption (TPE) has gained much attention and diverse study, for it has powerful capability of balancing the security and usability of cloud storage images by simultaneously securing privacy and preserving visual information of images. Unfortunately, most of the existing TPE schemes for JPEG images have the disadvantages of limited thumbnail precision or weak security or irreversibility, making them vulnerable to cryptanalysis. To address these issues, we propose a TPE based on adjustable precision (TPE-AP) for JPEG images. First, a joint adjustment strategy is introduced for encrypted quantized QC coefficient (QDCC) and quantized AC coefficient (QACC) of plain image, which makes the security and usability of the thumbnail controllable by exploiting numerical characteristics of QDCC and distribution features of QACC. Second, a time-varying encryption strategy based on international time is presented, characteristics of JPEG compression and quantized coefficients are combined to generate encryption sequences, improving the security of the whole encryption process. In addition, we explore the redundancy of QACC within DCT blocks and provide an improved embedding strategy to solve irreversibility problem. Experimental results show that the peak signal-to-noise ratio (PSNR) of thumbnail-preserving accuracy reaches 52 dB, the file expansion is minimally limited to 6.3%, the decryption time less than 0.13 s, and the mean average precision (mAP) of retrieval attains 62%, it indicates that TPE-AP outperforms the state-of-the-art methods.

Abstract The Siamese network-based tracker calculates object templates and search images independently, and the template features are not updated online when performing object tracking. Adapting to interference scenarios with performance-guaranteed tracking accuracy when background clutter, illumination variation or partial occlusion occurs in the search area is a challenging task. To effectively address the issue with the abovementioned interference and to improve location accuracy, this paper devises a Siamese residual attentional aggregation network framework for self-adaptive feature implicit updating. First, SiamRAAN introduces Self-RAAN into the backbone network by applying residual self-attention to extract effective objective features. Then, we introduce Cross-RAAN to update the template features online by focusing on the high-relevance parts in the feature extraction process of both the object template and search image. Finally, a multilevel feature fusion module is introduced to fuse the RAAN-enhanced feature information and improve the network’s ability to perceive key features. Extensive experiments conducted on benchmark datasets (GOT-10K, LaSOT, OTB-50, OTB-100 and UAV123) demonstrated that our SiamRAAN delivers excellent performance and runs at 51 FPS in various challenging object tracking tasks. Code is available at https://github.com/MallowYi/SiamRAAN .

Privacy-preserving localization plays a crucial role in enabling various applications on the Internet of Things (IoT). Existing work applies random zero-sum noise to develop privacy-preserving localization, which achieves efficiency and accuracy by adding random noise to preserve private information and cancelling the effect of the noise with the zero-sum characteristic, respectively. However, in practice, some nodes in IoT scenarios may misbehave, not following a pre-defined protocol but adding false noise or tampering with information, which leads to the trust issue for privacy-preserving localization. In this paper, we integrate blockchains with zero-sum noise to achieve trusted privacy-preserving localization against misbehaving nodes. Specifically, a three-layer framework is designed by combining private blockchains with a zero-sum noise-adding mechanism. In the sensing layer, nodes are divided into groups to perform the first noise-adding process to preserve their private location information during location aggregation inside the group. In the blockchain layer, each group constructs the private blockchains to achieve trustworthiness without increasing the risk of privacy leakage and performs the second noise-adding process to protect intermediate information. In the application layer, the target node aggregates the intermediate information from each group to estimate its location. Then, under the framework, we propose an Enhanced Privacy-Preserving Localization (EPPL) algorithm to securely calculate the location of the target node against misbehaving nodes. The correctness, accuracy, privacy, trustworthiness, and efficiency of EPPL are analyzed. The performance of EPPL is evaluated by using simulations. Compared with existing random noise-based methods, it is shown that EPPL can effectively enhance privacy-preserving localization.

Federated Learning (FL) offers a distributed machine learning framework that enables collaborative model training across multiple data sources without the need to share raw data, thereby preserving data privacy. This framework is particularly well-suited for cross-departmental and cross-enterprise intelligent decision-making in smart manufacturing. However, challenges remain in selecting reliable participants and ensuring the secure transmission of parameters to defend against potential attacks. Malicious participants may upload low-quality data or compromise data privacy during model aggregation. To address these issues, we propose the Federated Reputation Evaluation Blockchain (FREB), which integrates a reputation evaluation mechanism with blockchain technology. By leveraging blockchain, FL tasks are executed through trusted transactions, with smart contracts ensuring transparency and accountability. In contrast to traditional contribution evaluation methods, FREB employs a multi-weight subjective logic model combined with Shapley values to assess participant reliability. Reputation scores are calculated based on factors such as activity, model contribution, stability, and data quality, guiding the selection of participants. Additionally, a PoR-based model aggregation method is implemented, and noise is added to the model parameters to protect sensitive data from potential attacks. Experimental results on real-world datasets demonstrate that FREB effectively mitigates malicious node attacks and encourages high-quality participants, while maintaining model accuracy and data privacy.

Addressing expanded surveillance scopes, cross-device person re-identification emerges as a crucial issue. Traditional approaches to re-identification and tracking confront challenges in feature matching challenges, and fail to harness the potential of edge computing environments, which offer viable solutions to bandwidth and latency constraints. The paper introduces a novel, edge computing and shared cache-based method for person re-identification and tracking. In this framework, edge nodes exploit deep learning method to execute image recognition for person re-identification and trajectory tracking. Edge nodes are deployed within a tri-layered architecture, encompassing Hierarchical Edge Cloud (HEC), Mobile Edge Computing (MEC), and Central Cloud (CC). Emphasizing part feature significance for person re-identification, the approach employs a refined Part-based Convolutional Baseline model guided by the triplet loss function. Experimental outcomes validate the high re-identification precision and efficacious load balancing capability of this method in person re-identification and tracking operations. Achieving a harmonized trade-off between task accuracy and response time, the proposed framework offers a robust solution for person re-identification and tracking within edge computing contexts.

Clinical decision-making models have been developed to support therapeutic interventions based on medical data from either a single hospital or multiple hospitals. However, models based on multihospital data require collaboration among hospitals to integrate local data, which can result in information leakage and violate patient privacy. To address this challenge, we propose a novel approach that combines federated learning (FL) with inverse reinforcement learning (IRL) to create an efficient medical decision-making support tool while preserving patient privacy. Our approach uses an IRL algorithm with differential privacy to train a neural network-based agent on local data containing clinician trajectories, which learns a private treatment policy by observing patients’ conditions. Additionally, we integrate FL into the proposed algorithm to learn a global optimal action policy collaboratively among various smart intensive care units, overcoming data limitations at each hospital. We evaluate our approach using real-world medical data and demonstrate that it achieves superior performance in a distributed manner.

Federated learning (FL) provides an effective solution for multiparty data processing under privacy preserving, and becomes a good choice for crowd intelligence extraction in Health CrowdSensing. The quality of the local model submitted by the data holder determines the quality of the global model in FL, and the quality of the local model depends on the data quantity, data quality, and computing power of the data holder. However, in the process of model training, the data holder will inevitably spend the cost of communication and local model training, and higher quality data acquisition and higher quality local model training require higher cost. Therefore, how to motivate data holders with a large amount of high-quality data and computing power to participate in FL has become an urgent problem to be solved. This article transforms the problem of motivating data holders into an optimization problem of utility from the perspective of maximizing the utility of the data holder, establishes the incentive mechanism based on the Contract Theory, and proves that the optimal strategy set of the data holders reaches Nash Equilibrium. A large number of experiments based on public data sets of UCI and MNIST verify that the incentive mechanism can make the baseline algorithm converge faster, while resisting malicious behaviors, such as free-riding and collusive attacks. Furthermore, the data holder with a large amount of high-quality data and computing power can obtain higher revenue.

With the rapid development of embedded smart devices, a new data collection paradigm, mobile crowd-sensing (MCS), has been proposed. MCS allows individuals from the crowd to act as sensors and contribute their observation data. However, existing MCS systems are mostly based on third-party platforms, and there is no guarantee that a center is completely credible. In addition, security and privacy issues should not be ignored. During MCS’ execution, the participants’ various information and truth value are usually exposed, and the computation related to data privacy cannot be verified. In this paper, we integrate the blockchain into the MCS scenario to design a blockchain based privacy-preserving quality control mechanism, which prevents data from being tampered with, and denied, ensuring that the reward is distributed fairly. In the new system, we propose a privacy preserving participant selection scheme and the result can be verified (i.e., security against malicious node) without any third-party arbiter. Finally, considering the issues with sensing data privacy and efficiency in the truth discovery process, we propose a new privacy-aware crowdsensing design with iterative truth discovery based on rational secure multi-party computation. The experimental results show that compared to the prior result, the proposed solutions are highly practical and facilitate quality control without violating the participant’s privacy.

Recently with the popularity of embedded smart devices, mobile crowdsensing has become a new data collection paradigm. However, these traditional schemes rely on a central authority, leading to potential privacy disclosure, and there is no the guarantee that a center is completely credible in reality. In addition, the issues of security and privacy caused by the center platform should not be ignored, and the computation related to privacy data cannot be verified. In the paper, we propose a decentralized privacy-preserving model(PPNS) based on blockchain, which avoids data from being tampered with, and denied. In the prototype of PPNS, a privacy preserving node selection scheme is constructed and the result can be verified (i.e. security against malicious node) without any third-party arbiter. The experimental results show that compared with prior methods, our solution has a better performance and will facilitate node selection without violating participants' privacy.

In recent years, with the popularity of smartphones, the indoor positioning systems based on mobile crowdsensing (MCS) have gained considerable interest and exploit. However, it is still challenging to construct a largescale indoor positioning system. 1) In indoor positioning model, storage and computing resources are very important. 2) The calibration operation of data label and selection of model parameters require the operation of professionals. 3) User location privacy may be compromise, which greatly affects participant safety and enthusiasm. To solve these problems, our model firstly provides an edge-crowdsourcing indoor localization architecture to improve storage, computing power and response speed. Then, based on manifold regularization, a semi-supervised indoor localization model is determined by an adaptive manner in terms of both similarity and manifold structure, which reduces the workload of the positioning model and improve localization accuracy. In addition, we propose a new privacy-aware indoor localization algorithm based on secure multi-party computation to protect location privacy. Experimental results on real-world datasets show that, compared with the previous methods, our method improves accuracy by 0.87m, and in terms of time overhead of privacy protection, our method reduces the running time of the thousand seconds level.

With the popularization of intelligent terminals, crowdsensing has become increasingly prominent because of its advantages, such as low cost, high convenience, and fast speed in conducting tasks. However, the quality of the data collected through crowdsensing is varied and is difficult to evaluate. Furthermore, the existing crowdsensing quality control methods are mostly based on a central platform, which is not completely trusted in reality and results in the existence of fraud and other problems. To solve these two questions, a crowdsensing quality control model based on a two-consensus blockchain is proposed in this paper. First, the idea of a blockchain is introduced into this model. The credit-based verifier selection mechanism and the two-consensus approach are proposed to realize the nonrepudiation and nontampering of information in crowdsensing. Then, to help task publishers obtain higher-quality sensing data, the methods of node matching and QGE are proposed. The former method uses the idea of the calculation of matching degree to select workers, and the latter uses the idea of clustering and fuzzy theories to evaluate the quality of the sensing data. Finally, the experiments show that the running time of the block generation in our model is acceptable, and comparing with the other methods, our model can acquire data of higher ioj.

As mobile crowd sensing (MCS) cannot provide reliable services, the quality of service (QoS) is a major research interest. Since the service node makes a decisive impact on two vital factors of QoS, service node selection is becoming a novel research direction in MCS. In this paper, we analyze the factors that need attention when selecting proper service nodes in MCS and define the service node selection problem (SNSP) as follows: finding the optimal set of service nodes, provided that optimizes multiple metrics of QoS simultaneously and satisfies the network resource constraint. Accordingly, we formulate a multiobjective optimization model (MOOM), which converts SNSP to a multiobjective optimization problem (MOOP). Since the MOOM considers the comprehensive effect of all service nodes on one metric as one objective of MOOP, it can handle the diversity of metrics and conflicts between them; in particular, it can flexibly change the metric system of QoS depending on different demands. To demonstrate the value and effectiveness of the proposed MOOM, we propose a paradigm of it and design a corresponding multiobjective optimization selection mechanism. This paradigm focuses on the influence of node spatiotemporal mobility on both data collection and data transmission. Extensive experiments and comparison on a real-world data show that MOOM is an effective model for selecting service nodes with both good coverage and transmission performances.

As a mobile operating system framework, Android plays a significant role in supporting mobile apps. Using cloud-based approaches have further enabled the implementations of mobile apps. However, current Android application model is not efficient by using current common approaches. In this paper, we proposed a new mobile app model using precache technology to overcome the current existing obstacles. The results of experiments show that our model can reduce network traffic of Android apps efficiently. At the same time, using our proposed model can reduce the data redundancy and improve maintainability of Android apps.

With the powerful sensing capability of mobile smart devices, users can easily obtained the crowd sensing services with smart devices in the Internet of Things (IoT). However, credible interaction issues between mobile users are still the hard problems in the past. In this paper, we focus on how to assign the crowdsourcing sensing tasks based on the credible interaction between users. First, a novel credible crowdsourcing assignment model is proposed based on social relationship cognition and community detection. Second, the service quality factor (SQF), link reliability factor (LRF), and region heat factor (RHF) are introduced to scientifically evaluate the user crowdsourcing preferences. Then, a crowdsourcing algorithm based on analytic hierarchy process (AHP) theory is proposed. Finally, the simulation experiments prove the correctness, effectiveness, and robustness of our method.

The conventional security mechanism based on cryptography and authentication is not sufficient for countering all types of security problems resulting from internal malicious nodes and wrong act resulting from hardware breakdown and resources loss. The construction of a reputation based trustworthy sensor networks has recently become a hot point of making up for the shortcoming of conventional security mechanism. Firstly, a familiar reputation-based framework for sensor networks (RFSN) is presented, and problems caused by trust computing in the unreliable wireless channel are analyzed. Secondly, A Heider-theory based reputation framework for sensor networks (HRFSN) is presented for improving on RFSN using Heider theory. A node communicating with another node is observed by the mutual neighbor nodes of them, optimizing the calculation of direct reputation value when the channel is unreliable and enhancing the reliability of trust value. Finally, according to simulation analysis, the HRFSN framework can get more exact direct reputation value and dependable trust value than the RFSN framework. At the same time, this framework can maintain its resiliency against bad mouthing attacks and collusion attacks toward the recommended results.