Mobile crowdsensing (MCS), a human-in-the-loop IoT sensing paradigm, relies on active user participation. However, in untrusted environments, incentive design faces a fundamental trilemma among strict local differential privacy (LDP), individual rationality (IR), and system efficiency under limited budgets. LDP perturbation can obscure cost–quality correlations and undermine quality-aware selection, and it may even push selected users into negative utility. Meanwhile, most existing mechanisms either overlook privacy concerns or depend on heavyweight cryptographic frameworks, which are often impractical for resource-constrained mobile devices. To address these challenges, we propose IR-LDP, a lightweight LDP auction mechanism that explicitly manages the economic risk introduced by LDP noise under fixed budgets while guaranteeing deterministic IR. Specifically, IR-LDP introduces two design notions for LDP auctions: a public safety threshold that ensures ex-post IR for selected low-cost users, and a publicly bounded loss guarantee that caps each winner’s worst-case utility loss using only public system parameters. To realize these guarantees, IR-LDP employs a public projection strategy that projects noisy bids onto the safety threshold to bound downside utility risk, together with a nonlinear quality-aware ranking rule that prioritizes high-ROI users and mitigates the “lemon market” effect induced by LDP noise. We prove that IR-LDP satisfies ϵ-LDP for reported bids, maintains strict budget feasibility, and runs inO(NlogN) time. Extensive simulations across diverse budget regimes and user populations further demonstrate that IR-LDP outperforms representative baselines in social welfare, sensing quality, and budget efficiency, while providing ex-post IR for users with costs below the public threshold and a publicly known bound on utility loss for all participants.

Task complexity is widely regarded as a major barrier to cooperation in mobile crowdsensing (MCS), often leading to trust collapse and market failure. However, this view overlooks the constructive role of task complexity in shaping cooperative evolution. In this article, we propose a three-party evolutionary game framework involving workers, platforms, and task requesters, in which task complexity is explicitly modeled as an endogenous driver of trust dynamics and strategic interactions. We derive a set of anti-collapse conditions under which the marginal benefits of cooperative behavior overcompensate for the marginal costs induced by task complexity. Task complexity thereby propels an evolutionary phase transition from a low-trust trap to a stable cooperative equilibrium by reshaping the payoff structure of cooperative strategies. We further characterize the critical complexity thresholds that govern this phase transition through theoretical stability analysis. Extensive numerical simulations validate the theoretical predictions and demonstrate the robustness and effectiveness of the proposed mechanism in sustaining cooperative behavior across a wide range of task complexities.

Recently, adopting mobile crowdsensing to collect data, analyze information, and share knowledge has gained increasing attention from the research community. However, in practice, strategic selfishness and privacy breaches lead to user reluctance to participate in sensing tasks. To tackle the problem, we combine a two-way auction model with differential privacy to incentivize user participation in sensing activities while ensuring the protection of their sensitive information. In order to improve the match between users and tasks, we fully consider users' historical behavior and task attributes during the auction process, and reward users for completing tasks based on their performance. Furthermore, Laplace noise is added to users' sensitive information based on differential privacy to prevent privacy leakage. Through extensive simulations with real-world system settings, we verify that the proposed algorithms outperform other algorithms. Furthermore, we validate the soundness of modeling and the accuracy of analysis, which also guides the design of a more sophisticated incentive mechanism.

By locating computational and storage resources at the edge of the network, the emerging paradigm of Mobile Edge Computing (MEC) yields a significant enhancement in user Quality of Experience (QoE). However, the limited resources at edge nodes, coupled with the dynamism of user requests, present a considerable challenge to decision-making in service placement and computational resource assignment. This paper investigates the resource management problem within an edge-cloud cooperative network. Aiming to minimize long-term network latency, the original problem is first modeled as a Markov Decision Process (MDP) featuring a hybrid discrete-continuous action space. To address dynamically arriving tasks and varying network conditions, we develop a Dynamic Service Placement and Computation Resource Allocation (DSPCRA) scheme based on deep reinforcement learning (DRL). DSPCRA integrates a deep deterministic policy gradient (DDPG) with a parameterized action mechanism for online decision-making. Numerous simulations confirm that the proposed scheme exhibits good convergence properties and achieves lower latency performance compared to the benchmark algorithms.

Utilizing mobile crowdsensing (MCS) for data collection and analysis has become a prominent paradigm in the Internet of Things (IoTs). However, the existing research predominantly focuses on platform-user interactions, often neglecting the potential for user collaboration, which is crucial for improving data quality and task efficiency. In practical applications, mobile users tend to cooperate with familiar individuals based on their preferences in sensing tasks. To tackle this issue, we introduce a novel MCS model that integrates user cooperation, significantly enhancing the system's overall effectiveness. Specifically, users’ capabilities and costs are synthesized and managed through a cooperation degree matrix. Additionally, cooperation is updated based on historical behaviors and user preferences. To incentivize user participation, currencies are employed for recruitment. Within this framework, we investigate the maximum collaborative user selection (MCUS) problem, which is dedicated to the problem of maximizing the amount of recruitment under user cooperation. The MCUS problem is proved to be an NP-hard problem and thus intractable. To address this, we propose the minimum weighted cost replacement (MWCR) algorithm. Experimental results demonstrate that the MWCR algorithm exhibits low complexity and high efficiency across various scales, making it an excellent solution for collaborative crowd recruitment.

Graph kernels are a significant class of tools for measuring the similarity of graph data, which is the basis of a wide range of graph learning methods. However, graph kernels often suffer from high computing overhead. With the shining of cloud computing, it is desirable to transfer the computing burden to the server with abundant computing resources to reduce the cost of local machines. Nonetheless, under the honest-but-curious cloud assumption, the server may peek at the data, raising privacy concerns. To eliminate the risk of data privacy leakage, we propose CloudRGK to securely perform Random walk Graph Kernel(RGK), one of the most well-known graph kernels, on the cloud. We first prove that the edge- and vertex-labeled graphs could be transformed into an equivalent matrix representation. Afterward, we prove that the cloud could perform the core operations in RGK on the encrypted graphs without feature information loss. Evaluations of the real-world graph data demonstrate that our strategy significantly reduces the overhead of the local party to perform RGK without performance degradation. Meanwhile, it introduces only a small amount of extra computation cost. To the best of our knowledge, it is the first work towards private graph kernel computation on the cloud.

With the rapid growth of smart devices, mobile crowdsensing is becoming one of the most important and attractive paradigms to acquire information from physical environments. Low-quality data, a notorious but widely found issue, degrades the availability and preciseness of sensing services, especially for these complex sensing task scenarios. However, few existing incentive mechanisms frequently ignore the issue of data quality. In this paper, we define user reputation and user task preferences in a new perspective, while predicting the number of users likely to upload high-quality data by combining Poisson distribution. Then, the maximum expectation algorithm is employed to evaluate the parameter values of the Poisson distribution. Subsequently, a two-way auction mechanism is proposed, which encourages users to participate in the sensing task and improves the match between tasks and users. We adopt the number of high-quality data that the user may upload as a factor in the user’s offer to maximize the quality of data received by the platform. The analysis based on the model lays a theoretical foundation on the incentive process of mobile crowdsensing considering data quality. The evaluation results show that our mechanism outperforms other existing techniques, in terms of robustness and efficiency.

Caching content at the edge network has emerged as a critical technique to alleviate backhaul congestion, minimize service latency, and improve user Quality of Experience (QoE). Driven by operational profitability, the edge service provider (ESP) necessitates access to user preference data to optimize its caching policies. However, disseminating such sensitive information raises significant privacy concerns. To address this challenge, we propose a privacy-preserving cooperative edge caching framework that jointly enhances caching efficiency and safeguards user preference privacy. Specifically, we design a privacy-oriented popularity estimation protocol, PSRSA, which guarantees the preservation of user privacy while delivering precise content popularity estimations. The PSRSA protocol integrates a randomize-then-shuffle mechanism to obfuscate user data locally prior to aggregation and accommodates heterogeneous privacy preservation requirements across users by dynamically adjusting differential privacy (DP) budgets. Subsequently, we formulate a Stackelberg Game-based Edge Caching (SGEC) algorithm that jointly optimizes the utility functions of both the content provider (CP) and the ESP, thereby ensuring efficient resource allocation at edge servers. Comprehensive evaluations conducted on real-world datasets demonstrate the superiority of the integrated PSRSA+SGEC framework, exhibiting a 35.22% increase in caching utility and a 3.4% improvement in cache hit ratio, while simultaneously achieving a 20.1% reduction in privacy leakage risks, compared to state-of-the-art baselines.

Exploiting mobile crowdsensing to conduct data collection and analysis brings unprecedented opportunities to promote the development of the Internet of Things(IoT). However, malicious platforms may provide untrusted data or illegally leak users’ information, which leads users in crowdsensing networks to be reluctant to participate in sensing activities. Besides, users are unwilling to report malicious platforms without sufficient incentives. To tackle the problem, a new incentive mechanism is proposed by modeling users’ preferences in this paper. Specifically, two scenarios are considered to detect malicious platforms when users join sensing activities according to the system grasps user’s information, i.e., complete information scenario and partial information scenario. Different incentive algorithms are designed for each scenario to optimize the systems incentive cost. In the complete information scenario, we minimize the total incentive cost by ranking users’ preferences. In the partial information scenario, uniform Distribution and Laplace Distribution are employed to model the distribution of users’ preferences to find the optimal cost. Specifically, we incorporate the concept of non-convexity into design the incentive mechanism, when user preferences obey the Laplace Distribution. By conducting an in-depth exploration the properties of Laplace Distribution, we can transform it into a convex problem to solve it efficiently. The analysis based on these mechanisms lays a theoretical foundation on the detection of malicious platforms. Furthermore, the soundness of modeling and the accuracy of analysis are verified through extensive simulation, which also guides the design of more sophisticated incentive schemes for the detection of malicious platforms.

Encrypted communications, implemented for the confidential information exchange, facilitate the preservation of individual privacy. Unfortunately, some criminals abuse encrypted communications to conduct illegal activities, leading to the proliferation of the Darknet. To curb malicious darknet activities, the accurate and effective classification of darknet traffic is imperative. Considerable endeavors have been devoted to identifying the darknet traffic. However, the classification of darknet applications has not yielded a satisfactory result. This deficiency arises from the limitations of current approaches, e.g., some traditional methods rely on hand-crafted features that consume labor, and other neural network-based methods disregard the graph structure of the traffic. To tackle these challenges, we propose the Darknet Traffic Graph (DTG), a graph structure that captures the interactions between local clients and remote servers in darknet traffic. Furthermore, based on DTG, we combine the GNN model and attention mechanism to create the Darknet Graph Neural Networks, i.e., DGNN, a powerful model that sufficiently exploits the benign and darknet traffic features. As a result, on the CIC-Darknet2020 dataset, the accuracy of DGNN in traffic classification and application classification is 98.52% and 99.06%, respectively, which outperforms other classifiers.

Recently, adopting mobile-edge computing (MEC) to accommodate the compute-intensive and delay-sensitive tasks from mobile devices has gained increasing attention from the research community. In contrast to a cloud-centric scheme, deploying servers at the network edge offers the advantage of delivering faster and more efficient services. However, pioneering works primarily focus on a homogeneous server deployment strategy, which distributes the same quantity of servers among a specific number of selected locations. In this work, we aim to lay the theoretical foundation for budget-constrained profits maximization (BCPM) problem, which is a coupled problem of server deployment and task scheduling. Subsequently, a two-step optimization method is proposed. Through seeking the maximum matches in the constructed bipartite graph, a task scheduling algorithm is first designed to maximize the profits under the server deployment. Then, two approximation algorithms with provable approximation ratios are exploited to perform nearly optimal deployment of servers in a homogeneous and heterogeneous manner, respectively. Extensive simulations with real-world data set and system settings are conducted. The results show that the proposed algorithms can achieve at least a 10.54% increase in total profits and the average processing delay of tasks can be shortened by about 17%.