With the rapid advancement of technology, mobile crowdsensing (MCS) has become a key enabler of improved daily life, drawing on its unique advantages. However, MCS systems face significant challenges concerning location privacy leakage, particularly during task allocation. To address the risk of worker location privacy leakage, we propose a location privacy-preserving system for crowdsensing based on secure sorting (LPPCS). LPPCS integrates elliptic curve cryptography (ECC) and secure multi-party computation (SMPC) technologies. By sorting the actual distances between workers and task locations, it achieves precise task assignment while ensuring the confidentiality and integrity of workers’ location data. The system consists of two main parts. The first part introduces a reverse sealed-bid auction mechanism, which reduces the computational costs by scaling down the number of workers participating in subsequent encryption computations. The second part designs a secure sorting protocolSSP. This protocol ensures that each worker only learns the relative sorting of their true distance to the task location and uploads this information to the server for final task assignment. Crucially, workers cannot access information about one another. This not only effectively protects their location privacy but also improves the accuracy of task assignments—all without the need for a trusted server. Finally, we conducted comprehensive evaluations using both simulated and real-world datasets. The results show that LPPCS performs exceptionally well in reducing computational costs, protecting workers’ location privacy, and improving task allocation precision.

The rapid growth of IoT devices has led to an increase in computation-intensive and latency-sensitive tasks, making traditional cloud computing insufficient. Cloud-edge-terminal collaboration can enhance offload efficiency by optimizing processing latency, energy consumption, and price cost. However, in real-world networks, computational results often need to be transmitted to multiple users, increasing the offloading complexity. This paper proposes a three-tier collaborative offloading architecture for interacting users, considering constraints such as service caching and various types of resources. The optimization problem of computation latency and price cost is modeled as a Markov Decision Process. To address the problem, we propose a deep reinforcement learning algorithm based on the soft actor-critic framework. Given the discrete-continuous hybrid action space, the algorithm incorporates a dual-head mechanism. After that, transfer learning is incorporated into the training strategy to improve adaptability in dynamic environments. The simulation results demonstrate that the proposed approach outperforms existing performance, convergence, and adaptability methods.

Abstract Mobile Crowdsensing (MCS) is considered to be a key emerging example of a smart city, which combines the wisdom of dynamic people with mobile devices to provide distributed, ubiquitous services and applications. In MCS, each worker tends to complete as many tasks as possible within the limited idle time to obtain higher income, while completing a task may require the worker to move to the specific location of the task and perform continuous sensing. Thus the time and location information of each worker is necessary for an efficient task allocation mechanism. However, submitting the time and location information of the workers to the system raises several privacy concerns, making it significant to protect both the temporal and spatial privacy of workers in MCS. In this article, we propose the Task Allocation with Temporal‐Spatial Privacy Protection (TASP) problem, aiming to maximize the total worker income to further improve the workers' motivation in executing tasks and the platform's utility, which is proved to be NP‐hard. We adopt differential privacy technology to introduce Laplace noise into the location and time information of workers, after which we propose the Improved Genetic Algorithm (SPGA) and the Clone‐Enhanced Genetic Algorithm (SPCGA), to solve the TASP problem. Experimental results on two real‐world datasets verify the effectiveness of the proposed SPGA and SPCGA with the required personalized privacy protection.

Due to the swift advancement of edge computing and mobile crowdsensing (MCS), edge-assisted MCS (EAMCS) has emerged as a promising paradigm, leveraging sensor-embedded mobile devices for the collection and sharing of environmental data. As the sensing scale increases in the modern urban, the application scenario becomes more and more complex, and the budget of users and platform is limited. Therefore, it is indispensable to study the effective task allocation mechanism with considering the multiple budget constraints in the EAMCS system. However, a majority of the existing studies unilaterally focus on either the users’ time budget or the platform’s budget, disregarding the crucial aspect of the users’ energy budget. In this article, we design a joint user movement, sensing, offloading, and computation framework adopting the computation offloading strategy called binary processing strategy. In addition, the multitask data collection with a limited budget (MDCB) problem considering time, energy, and platform budget in EAMCS is formulated, which is proved to be nondeterministic polynomial-hard. In order to maximize the amount of data collected by the users in the MDCB problem, we first verify the submodularity of the objective function, then propose the global maximum data first search algorithm and task sequence-based genetic algorithm to solve the problem. The extensive experiments are conducted on both synthetic and real-world data sets to demonstrate the effectiveness of our proposed schemes.

In recent years, wireless rechargeable sensor networks (WRSNs), as a crucial technology in cyber–physical–social systems (CPSSs), have gradually become a hotspot of research, with the development of wireless energy transmission technology. In previous works, the objective is to maximize the survival rate of sensor nodes. However, in this article, we focus on maintaining more targets. First, it details the charging scheduling problem of maximizing coverage of targets (CoT) in on-demand charging architecture of WRSNs. Also, the problem is formalized as a multiple-objective optimization problem, which aims at maximizing the CoT and the energy efficiency simultaneously. After that, the charging scheduling for maximizing coverage of targets (CSCT) scheme is proposed to achieve the above objectives. Then, the problem is reformulated as a Deadline-TSP problem that is NP-hard. To address this problem, we design an energy predictive model and propose the CSCT with an$n$-path ($n$-CSCT) scheme that has an$O(|\mathcal{N}|^n)$computational complexity. In addition, the resurrection of sensor nodes is considered in this article. Thus, the$n$-CSCT with node resurrection ($n$-CSCT-R) scheme is proposed for this case. Finally, we validate the effectiveness of the proposed schemes via extensive simulations.

Over the last decade, the Internet of Things (IoT) technology has advanced significantly in a variety of fields. As a pivotal application of IoT, intelligent transportation systems (ITS) have harvested great attention from the research community. Radio frequency identification (RFID) which is an essential technology in IoT plays a key role in ITS to identify tagged vehicles. Unknown tag identification which aims at identifying the existing unknown tags is crucial to monitor the newly entering vehicles in the RFID-assisted intelligent transportation systems. However, the COTS (commercial-off-the-shelf) RFID tags that harvest energy from the reader can not support the hash function in reality, which hinders the widespread deployment of hash-enabled unknown tag identification protocols. To conquer this tough issue, we propose two approaches to efficiently identify unknown COTS RFID tags. We first propose a Single-Point Selective unknown tag identification approach called SPS, where an analog hash pattern using the EPC (Electronic Product Code) segments is deployed to exclusively identify unknown tags. An unknown tag will be identified when it selects a singleton slot to reply. To improve the time efficiency of SPS, we further propose a Multi-Point Selective unknown tag identification approach called MPS. In MPS, two techniques of batch identification and batch division are developed to reduce the number of empty slots and avoid tag collisions, respectively. Then the parameters are theoretically analyzed to maximize the identification efficiency. The effectiveness of the proposed approaches is validated via both the simulations and COTS RFID device based experiments.

In recent years, wireless rechargeable sensor networks (WRSNs) have gained significant attention in the research community due to the current advancements in wireless power transfer technology. In mobile charger scheduling, previous works primarily emphasized the survival rate of sensor nodes. However, the primary task of a WRSN is to monitor targets in a given area. Therefore, the coverage of targets (CoT) maximization should be the primary objective of mobile charger scheduling. In this paper, we shift the focus to the CoT maximization on-demand charging scheduling problem, and formulate it as a multi-objective optimization problem, aiming to simultaneously enhance the average coverage and energy efficiency. We prove that the problem is NP-hard by reformulating it as a Multiple Travelling Salesman Problem with Deadline. We first propose the multiple chargers scheduling scheme for maximizing coverage of targets called MaxCov, which is designed to optimize the charging scheduling process and improve network performance in terms of coverage. Then, we further propose the multiple chargers scheduling scheme based on requests grouping called MaxCov-RG, which can well balance the trade-off between the performance and computational complexity. Finally, we validate the effectiveness of the proposed schemes via extensive simulations.

Federated learning (FL) is a distributed machine learning paradigm that addresses the challenges of privacy leakage and data silos by collaboratively training the global model through parameter exchange, rather than data, between the central server and local clients. However, recent researches highlight the vulnerability of FL to gradient leakage attacks where adversaries exploit shared parameters from clients to reconstruct sensitive training data. Differential privacy (DP) effectively mitigates this threat by adding noise to shared parameters, yet introduces a trade-off between privacy and accuracy in FL. To better balance the privacy and accuracy, in this paper we propose a staged noise perturbation strategy, called alternating noise permutation (ANP), from a novel perspective. ANP adds Gaussian-distributed random noise to model parameters during the critical learning period of FL, following DP principles. While in non-critical learning period, ANP alternately permutes the noise during odd and even communication rounds, achieving near mutual cancellation and mitigating the negative impact. Experimental results across three datasets and two neural networks under both independent identical distribution (IID) and NonIID scenarios demonstrate that ANP significantly improves classification accuracy and exhibits robustness against gradient leakage attack, ensuring the effectiveness of FL for secure and accurate collaborative model training.