Multimodal fake information increasingly threatens the Web ecosystem's trustworthiness and security, making improving detection accuracy a critical scientific challenge. The limited information interaction in traditional multimodal fake news detection methods fails to leverage semantic knowledge to model complex cross-modal forgery patterns and global structural anomalies, restricting the model's capability. To address the issues, this paper proposes a multimodal fake news detection method, SVPF-Net, that centers on semantic-driven visual enhancement and knowledge-aided modality-priority fusion. For visual representation optimization, we design a dual-feature extraction module and a dual-fusion enhancement module. A weighted fusion strategy is employed to construct a structured visual representation that integrates the semantics of local forgeries and global anomalies. Meanwhile, a cross-attention mechanism enables bidirectional alignment and interactive coupling between local and global image features, thereby achieving effective complementarity between local forgery cues and global anomaly patterns. For multimodal fusion, high-quality textual semantic features and visual representations are integrated via a modality-priority progressive fusion strategy that relies on cross-attention. The integration enables robust cross-modal semantic interaction and effectively enhances the efficiency of multimodal feature fusion. Comprehensive experiments validate the optimal performance of SVPF-Net and its ability to enhance interpretable semantics, providing valuable support for the practical application of reliable fake news detection.

The deep vision sensing has been a practical tool in early disease detection, and this work aims at an important branch of ocular disease recognition. Although a number of researchers had paid attention to it during past years, fine-grained ocular feature extraction always remains a challenge. To handle with this issue, this work benefits from comprehensive ability of the convolution-Transformer structure (Conformer), and proposes vision sensing-driven intelligent ocular disease detection using conformer-based dual fusion. On the one hand, the proposal combines technical advantages of convolution and visual Transformer to more accurately fuse local subtle features and global representation information in images. On the other hand, the proposal significantly improves accuracy and robustness of the model by optimizing depth and width. Simulation experiments on real-world ocular disease image datasets show that the proposed model exhibits higher performance in ocular disease detection compared to other methods. Numerical results show that it improves the detection accuracy by 1% to 3.7% compared to several mainstream baseline methods. This research result not only promotes the development of ocular disease detection, but also provides more reliable technical support for accurate diagnosis of ophthalmic diseases.

The rapid advancement of Internet of Things (IoT) and multi-agent systems has transformed how complex IoT tasks are managed across domains. While individual edge agents demonstrate proficiency in specialized tasks such as data collection and edge learning, they encounter substantial challenges when confronting complex IoT scenarios that demand diverse skill sets. This paper introduces a novel group formation framework facilitating effective IoT agent collaboration in complex task environments, including smart manufacturing, intelligent transportation, and smart cities. We propose a hybrid competition mechanism that optimizes initial multi-agent cooperation strategies by integrating task requirements, agent capabilities, and system-wide performance metrics. Our approach combines intra-task and inter-task competition to achieve optimal agent-task matching and resource allocation in large-scale IoT networks. Through comprehensive simulations across various IoT scenarios, we demonstrate that our framework substantially enhances task completion efficiency and system performance compared to existing methods. The results confirm our approach’s effectiveness in resource-constrained environments, achieving minimal agent grouping time costs while increasing total task revenue by 30%-42% and resource utilization by 38% compared to baseline heuristic methods.

The Industrial Internet of Things coupled complex networks (IIoT-CCNs) contain a large and diverse number of nodes, whose states are inherently random and complex. In this context, federated learning can play a significant role. It allows different nodes or subsets of the IIoT-CCNs to train local models without the need to transfer all their raw data to a central server. This paper firstly establishes a more general IIoT-CCNs model, where the communication channel may not be completely opened between coupled nodes. When integrating federated learning, the nodes can collaboratively train a global model while safeguarding their own data sovereignty. Secondly, a pinning impulsive controller is designed to make IIoT-CCNs realize synchronization. Thirdly, by utilizing regroup method and step-function methods, some useful and novel synchronization stability conditions have been obtained. In the federated learning framework, these stability conditions can be adapted to ensure the convergence of the collaborative learning process. Then, a simulation is conducted to make sure the correctness of results. Finally, the application of synchronization with regard to responsible image communication is executed. The feature values of node data can be extracted through the image encryption algorithm, and then using synchronized chaotic sequences obtained from encryption to ensure the security of image information in the transmission process. Even in a federated learning-enabled IIoT-CCNs, the encrypted data and synchronized sequences can be managed in a way that respects the privacy and security requirements. It also ensures that legitimate recipients can simultaneously obtain the correct keys according to the image encryption algorithm. By utilizing the histogram and adjacent pixel correlation analysis to verify the effectiveness of the encryption scheme.

In recent years, data-driven remote medical management has received much attention, especially in application of survival time forecasting. By monitoring the physical characteristics indexes of patients, intelligent algorithms can be deployed to implement efficient healthcare management. However, such pure medical data-driven scenes generally lack multimedia information, which brings challenge to analysis tasks. To deal with this issue, this paper introduces the idea of ensemble deep learning to enhance feature representation ability, thus enhancing knowledge discovery in remote healthcare management. Therefore, a multiview deep learning-based efficient medical data management framework for survival time forecasting is proposed in this paper, which is named as "MDL-MDM" for short. Firstly, basic monitoring data for body indexes of patients is encoded, which serves as the data foundation for forecasting tasks. Then, three different neural network models, convolution neural network, graph attention network, and graph convolution network, are selected to build a hybrid computing framework. Their combination can bring a multiview feature learning framework to realize an efficient medical data management framework. In addition, experiments are conducted on a realistic medical dataset about cancer patients in the US. Results show that the proposal can predict survival time with 1% to 2% reduction in prediction error.

In context of 6G mobile computing, the combination of Industrial Internet of Things (IoT) and machine learning extends intelligent sensing ability to improve industrial operation efficiency. In conventional operation of sewage treatment process (STP), manipulators often made excessive aeration amount in treatment process, in order to reach environmental standard. However, such rough operation mode will bring redundant energy consumption. To deal with this issue, this work employs industrial 6G-IoT environment provide basic data conditions for intelligent sensing scheme. On this basis, an industrial 6G-IoT sensing and machine-learning-supported intelligent sensing framework is established for indicator control strategy in STP. In particular, the amount of dissolved oxygen (DO) is selected as the main control object. Then, given inlet conditions and expected outlet conditions, the support vector regression model is formulated to predict the appropriate DO amount values. The proposed approach is evaluated on data collected from a real-world industrial 6G-IoT-based STP. And it is compared with several typical machine-learning-based prediction methods. Numerical results show that the method proposed is 5% better than the typical methods with a deviation of less than 0.6 and can achieved prediction precision about 80%.

Effective data management schemes have always been the major demand in universal industrial Internet of Things (IoT) systems, especially in resource-constrained scenarios. In realistic wastewater treatment process (WTP), only limited monitoring data resource can be available due to some digital constraint. Aiming at this practical issue, this work explores utilization of deep neural network to deal with such practical issue in the objective situation. Therefore, a deep learning-based data management scheme for intelligent control of WTP under resource-constrained IoT systems, is proposed in this paper. Firstly, a specific data encoding and preprocessing approach is developed for the objective business scenario. Then, the detailed workflow of a deep neural network structure is applied to predict key intermediate parameters which can further guide control decision. Finally, a comprehensive series of experiments are conducted on a real-world dataset which covers a range of one year. Both efficiency and robustness of the proposal are tested by introducing several performance metrics. The results show that it can have proper prediction effect in such resource-constrained environment, which can facilitate following intelligent control operations.

The finance-level Artificial Intelligence of Things (AIoT) is going to become a novel media in the 6G-driven digital society. Inside the financial AIoT environment, large-scale crowd credit assessment with the guarantee of low latency has been a general demand. Facing limited computational resources, there is still a lack of effective computation offloading methods for this purpose to ensure low latency. In order to deal with such an issue, this article introduces edge computing mode and proposes a low-latency edge computation offloading scheme for trust evaluation in financial AIoT. With different elements involved in the assessment process being denoted via mathematical description, a multiobjective optimization problem with constraints is formulated. Then, the aforementioned optimization problem is solved by a specific search algorithm, so that optimal task offloading schemes can be found. To assess the performance of the proposal, some simulation experiments are conducted to verify the proposed task offloading method. And it can be reflected from numerical results that latency can be well reduced compared with baseline methods.

Fake news detection has been a more urgent technical demand for operators of online social platforms, and the prevalence of deep learning well boosts its development. From the model structure, existing research works can be categorized into three types: convolution filtering-based neural network approaches, sequential analysis-based neural network approaches, and attention mechanism-based neural network approaches. However, almost all of them were developed oriented to scenes of a single language, without considering the context of mixed languages. To bridge such gap, this article extends to the basic pretraining language processing model transformer into the multiscale format and proposes a novel fake news detection model for the context of mixed languages through a multiscale transformer to fully capture the semantic information of the text. By extracting more fruitful feature levels of initial textual contents, it is expected to obtain more resilient feature spaces for the semantics characteristics of mixed languages. Finally, experiments are conducted on a postprocessed real-world dataset to illustrate the efficiency of the proposal by comparing performance with four baseline methods. The results obtained show that the proposed method has an accuracy of about 2%–10% higher than commonly used baseline models, indicating that the scheme has appropriate detection efficiency in mixed language scenarios.

Intelligent detection techniques for online spamming have been a hot concern in academia. Although much technical progress has been achieved in recent years, two aspects of challenges are still confronted by scholars. For one thing, spamming activities are accompanied by multisource attributes, such as behaviors and semantics. For another, spamming is a cross-platform activity, where multiple platforms are exploited simultaneously to expand the influential reach. The above circumstances actually make spamming detection tend to become a fuzzy detection task. Existing works typically consider one-sided attribute and lack cross-platform multifeature fusion, which limiting the effectiveness of detection. To handle the current challenges, this article proposes a cross-field deep learning-based fuzzy spamming detection approach via the collaboration of behavior modeling and sentiment analysis. First of all, a cross-field deep learning-based technical framework is put forward to implement multisource feature fusion from mixed context. It first extracts multisource features from single fields and then integrates them into a hybrid-field feature space. In addition, three cross-field datasets based on real-world social network datasets are constructed, and utilized in the evaluation of our proposed approach. The findings demonstrate that our proposal improves the detection accuracy by about 7% to 12%, in comparison to five other baseline approaches.

Existing infrared small target detection algorithms often lack adaptability in complex scenes and heavily rely on parameter configurations. To address this limitation, we propose a novel infrared small target detection method based on adaptive region growing algorithm with iterative threshold analysis that leverages the homogenous compactness of the small target and discontinuity with its surroundings. Initially, the image undergoes adaptive splitting into multiple regions using an automatic seeded region growing (ASRG) algorithm, eliminating the need for preassigned seed points. Next, the segmentation results at each threshold are utilized to calculate the relative residual map (RRM) and local dissimilarity map (LDM), contributing to the selection of the optimal threshold. Finally, RRM and LDM corresponding to the optimal threshold are integrated to accurately characterize the small target signal while effectively removing background clutter. Experimental results show that the proposed method is effective in clutter removal and small target detection in diverse complex scenes, and is robust to the shape and size of targets.

With the rapid development of edge intelligence in wireless communication networks, mobile-edge networks (MENs) have been broadly discussed in academia. Supported by considerable geographical data acquisition ability of mobile Internet of Things (IoT), the MENs can also provide spatial locations-based social service to users. Therefore, suggesting reasonable points-of-interest (POIs) to users is essential to improve user experience of MENs. As the simple user-location data is usually sparse and not informative, existing literature attempted to extend feature space from two perspectives: 1) contextual patterns and 2) semantic patterns. However, previous approaches mainly focused on internal features of users, yet ignoring latent external features among them. To address this challenge, in this article, a deep distributed-learning-based POI recommendation (Deep-PR) method is proposed for situations of MENs. In particular, hidden feature components from both local and global subspaces are deeply abstracted via representative learning schemes. Besides, propagation operations are embedded to iteratively reoptimize expressions of the feature space. The successive effect of the above two aspects contributes a lot to more fine-grained feature spaces, so that a recommendation accuracy can be ensured. Two types of experiments are also carried out on three real-world data sets to assess both efficiency and stability of the proposed Deep-PR. Compared with seven typical baselines with respect to four evaluation metrics, obtained results of the overall performance of the Deep-PR are excellent.

In future smart cities supported by cyber-physical social intelligence, autonomous behavioral decision for vehicular agents is going to become a general demand. Despite much progress achieved in autonomous behavioral decision of vehicular agents, the existing works can just be used in scenarios of short-distance behavioral decision. Naturally, they are not well suitable for long-distance behavioral decision tasks, posing much challenge in realistic cyber-physical environment. To bridge the existing gaps, this article proposes an autonomous behavioral decision framework for vehicular agents using cyber-physical social intelligence. First, it is expected to establish a dynamic planning model with multiple objectives and constraints. This can be embedded into the control unit of a vehicular agent to endow it with proper social intelligence. On this basis, an iterative search algorithm is specifically designed for it to find the optimal solutions from the whole solution space. Finally, two typical situation cases are implemented with use of simulation modeling to display the working architecture of the proposed method. In addition, a universal optimization search algorithm is selected as the baseline to be compared with the proposed method. The comparison results reveal both planning utility and running efficiency of the proposed method.

Currently, health management driven by intelligent means is a general demand of social systems. Although a number of researchers have paid attention to such areas, they have primarily focused on improving the performance of intelligent algorithms. Such intelligent algorithms are mostly based on the central computing mode, where all the user data are aggregated together in a central cloud to implement computing tasks. This poses a great threat to personal privacy due to exposure to the outside world. To address this challenge, this work uses a federated learning mechanism and proposes a privacy-preserving social computing framework for health management. User data are deposited in different user terminals to prevent exposure. A group of parameters are pretrained for each terminal in an iteration and are then transferred to the center cloud for updating. After multiple rounds of interactive training between the center cloud and the terminals, a recognition model finishes training for each terminal without direct access to data from other sources. Finally, this work also conducts experiments on a real-world dataset to assess the overall performance of the proposed approach.

The rapid development of the Internet of Vehicles has substantially boosted the prevalence of vehicular social networks (VSN). However, content security has gradually been a latent threat to the stable operation of VSN. The VSN is a time-varying environment and mixed with various real or fake contents, which brings great challenges to the sustainability of VSN. To establish a sustainable VSN, it is of practical value to possess a strong ability for fake content detection. Related works can be divided into the global semantics-based approaches and the local semantics-based approaches, though both with limitations. Leveraging these two different approaches, this paper proposes a fake content detection model based on the mixed graph neural networks (GNN) for sustainable VSN. It takes GNN as the bottom architecture and integrates both convolution neural networks and recurrent neural networks to capture two aspects of semantics. Such a mixed detection framework is expected to possess a better detection effect. A number of experiments were conducted on two social network datasets for evaluation, and the results indicated that the detection effect can be improved by about 5%-15% compared with baseline methods.

Intelligent algorithms-driven industrial decision systems have been a general demand for modeling complex sewage treatment processes (STP). Existing researches modeled complex STP with the use of various neural network models, yet neglecting the fact that latent and occasional relations exist inside complex STP. To deal with the challenge, this paper proposes graph embedding-based intelligent industrial decision for complex STP (GE-STP). The graph embedding (GE) scheme is employed to enhance feature extraction and neural computing structure is utilized to simulate uncertain biochemical transformation inside STP. The introduction of GE can not only improves the fineness of feature spaces, but also improves the representative ability of models towards complex industrial processes. On this basis, the GE-STP is evaluated on a real-world data set collected from a realistic sewage treatment plant equipped with a set of Internet of Things devices. And some typical neural network models that have been utilized for modeling complex STP, are selected as baseline methods. Three groups of experiments show that efficiency of the GE-STP exceeds baselines about 6%–12%, and that the GE-STP is not susceptible to parameter changing.

The explosion of intelligent mobile applications has generated an unprecedented increase in the demand for diverse social content access. Device-to-device (D2D) communication, which enables direct transmission of content, can partly support these needs by reusing cellular spectrum resources. Most existing works focus on mitigating interference in D2D communication only in physical space. However, the effect of user content preference on interference in social space is ignored. In this article, a satisfied matching-embedded Social Internet of Things (IoT) is proposed for content preference-aware resource allocation (SIoT-RA). The Social IoT architecture is explored for user social characteristic analysis scenarios to ensure credible data management. Specifically, the user’s content preference characteristics are modeled as the probability of selecting similar content in the social IoT by employing the Dirichlet process. A satisfied matching-based model is proposed and embedded into the architecture as the core algorithm to achieve content preference-aware resource allocation. This design not only helps to reduce interference from the perspective of social space but also realizes maximized and satisfactory resource allocation performance by a matching algorithm that can quantify satisfaction as perceived utility. Extensive simulation results demonstrate the significant performance gains and high cost effectiveness of the proposed core algorithm of SIoT-RA in terms of the weighted sum data rate and users matching satisfaction.

The emergence of Artificial Intelligence of Things (AIoT) has provided novel insights for many social computing applications, such as group recommender systems. As the distances between people have been greatly shortened, there has been more general demand for the provision of personalized services aimed at groups instead of individuals. The existing methods for capturing group-level preference features from individuals have mostly been established via aggregation and face two challenges: 1) secure data management workflows are absent and 2) implicit preference feedback is ignored. To tackle these current difficulties, this article proposes secure AIoT for implicit group recommendations (SAIoT-GRs). For the hardware module, a secure Internet of Things structure is developed as the bottom support platform. For the software module, a collaborative Bayesian network model and noncooperative game are introduced as algorithms. This secure AIoT architecture is able to maximize the advantages of the two modules. In addition, a large number of experiments are carried out to evaluate the performance of SAIoT-GR in terms of efficiency and robustness.

Nowadays, online spamming has already been a remarkable threat to contents security of Internet of Things. Due to constant technical progress, online spamming activities have been more and more concealed. This brings much fuzziness to spammer detection scenarios, yielding the issue of fuzzy detection of spammers. Although existing detection techniques for spammers utilized idea of deep learning, they still ignore to release power of label spaces. As real nature about a user may be usually fuzzy, but the label annotated for a user is always certain. To remedy such gap, this article proposes a label smoothing-based fuzzy detection method for spammers (Fuz-Spam). First of all, deep representation is still utilized to deeply fuse features, which acts as the foundation of neural computing. On this basis, generative adversarial learning is introduced to transform previous label spaces into distributed forms. In addition, two groups of experiments are carried out on two real-world datasets for evaluation. The results demonstrate that the Fuz-Spam improves identification efficiency about 10% to 20% than previous ones, and that the Fuz-Spam is endowed with proper stability.

In ear of smart cities, intelligent medical image recognition technique has become a promising way to solve remote patient diagnosis in IoMT. Although deep learning-based recognition approaches have received great development during the past decade, explainability always acts as a main obstacle to promote recognition approaches to higher levels. Because it is always hard to clearly grasp internal principles of deep learning models. In contrast, the conventional machine learning (CML)-based methods are well explainable, as they give relatively certain meanings to parameters. Motivated by the above view, this paper combines deep learning with the CML, and proposes a hybrid intelligence-driven medical image recognition framework in IoMT. On the one hand, the convolution neural network is utilized to extract deep and abstract features for initial images. On the other hand, the CML-based techniques are employed to reduce dimensions for extracted features and construct a strong classifier that output recognition results. A real dataset about pathologic myopia is selected to establish simulative scenario, in order to assess the proposed recognition framework. Results reveal that the proposal that improves recognition accuracy about two to three percent.

With the continuous progress of communication quality, the wireless vehicular networks (WVN) will surely be-come an inevitable part of future smart cities. Inside WVN where context is complicated and stochastic, quality of service (QoS) acts as the core concern for broad users. And reliable prediction towards traffic in WVN is essentially an important demand to ensure QoS. Conventionally, related methods mainly focus one side to establish robust prediction models, possessing some limitations. To bridge such gap, model integration may be an intuitive and promising solution. This paper proposes QoS-aware reliable traffic prediction model under WVN (TP-WVN). Firstly, two typical prediction models are used as fundamental learners, which can capture the spatial correlations from different angles. Then, regression model is selected as the integrator to combine base models together. Simulative experiments on a real-world dataset are conducted to evaluate the proposal, and results show that the TP-WVN is able to realize reliable QoS-aware prediction compared with baseline methods.

Spamming is emerging as a key threat to the Internet of Things (IoT)-based social media applications. It will pose serious security threats to the IoT cyberspace. To this end, artificial intelligence-based detection and identification techniques have been widely investigated. The literature works on IoT cyberspace can be categorized into two categories: 1) behavior pattern-based approaches and 2) semantic pattern-based approaches. However, they are unable to effectively handle concealed, complicated, and changing spamming activities, especially in the highly uncertain environment of the IoT. To address this challenge, in this article, we exploit the collaborative awareness of both patterns, and propose a Collaborative neural network-based spammer detection mechanism (Co-Spam) in social media applications. In particular, it introduces multisource information fusion by collaboratively encoding long-term behavioral and semantic patterns. Hence, a more comprehensive representation of the feature space can be captured for further spammer detection. Empirically, we implement a series of experiments on two real-world data sets under different scenarios and parameter settings. The efficiency of the proposed Co-Spam is compared with five baselines with respect to several evaluation metrics. The experimental results indicate that the Co-Spam has an average performance improvement of approximately 5% compared to the baselines.

To reduce carbon footprint, a hybrid energy powered cellular network (HybE-Net) in the Internet-of-Things (IoT) environment is widely sought after. Different from cellular network powered on-grid energy, the base station (BS) wakeup in HybE-Net needs to consider the solar energy of the BS and the traffic load in the network. Thus, in this article, a fuzzy logic-based wakeup strategy is proposed, which comprehensively considers the energy wakeup level and the available network resource ratio. Then, the solar energy states are analyzed mathematically by using the diffusion approximation method. Finally, to prevent BSs from switching frequently between the sleeping and nonsleeping model, the awakening threshold is optimized by the penalty function method. The simulation results demonstrate that in the proposed wakeup strategy, the wakeup threshold based on the energy state is used to avoid the phenomenon of frequent handoff of the BSs, resulting in fewer handoffs. The proposed BS wakeup strategy can be further applied to both the current and sixth-generation (6G) mobile communication networks, which will be powered by other forms of renewable energy and on-grid energy in the future.

Nowadays, rumor spreading has gradually evolved into a kind of organized behaviors, accompanied with strong uncertainty and fuzziness. However, existing fuzzy detection techniques for rumors focused their attention on supervised scenarios that require expert samples with labels for training. Thus, they are not able to well handle the unsupervised scenarios where labels are unavailable. To bridge such gap, this article proposed a fuzzy detection system for rumors through explainable adaptive learning. Specifically, its core is a graph embedding-based generative adversarial network (Graph-GAN) model. First of all, it constructs fine-grained feature spaces via graph-level encoding. Furthermore, it introduces continuous adversarial training between a generator and a discriminator for unsupervised decoding. The two-stage scheme not only solves the fuzzy rumor detection under unsupervised scenarios, but also improves robustness of the unsupervised training. Empirically, a set of experiments are carried out based on three real-world datasets. Compared with seven benchmark methods in terms of four metrics, the results of the Graph-GAN reveal a proper performance, which averagely exceeds baselines by 5–10%.

Nowadays, the issue of information overload is gradually gaining exposure in the Internet of Things (IoT), calling for more research on recommender system in advance for industrial IoT scenarios. With the ever-increasing prevalence of various social networks, social recommendations (SoR) will certainly become an integral application that provides more feasibly personalized information service for future IoT users. However, almost all of the existing research managed to explore and quantify correlations between user preferences and social relationships, while neglecting the correlations among item features which could further influence the topologies of some social groups. To tackle with this challenge, in this article, a deep graph neural network-based social recommendation framework (GNN-SoR) is proposed for future IoTs. First, user and item feature spaces are abstracted as two graph networks and respectively encoded via the graph neural network method. Next, two encoded spaces are embedded into two latent factors of matrix factorization to complete missing rating values in a user-item rating matrix. Finally, a large amount of experiments are conducted on three real-world data sets to verify the efficiency and stability of the proposed GNN-SoR.

With the prevalence of Internet of Things (IoT)-based social media applications, the distance among people has been greatly shortened. As a result, recommender systems in IoT-based social media need to be developed oriented to groups of users rather than individual users. However, existing methods were highly dependent on explicit preference feedbacks, ignoring scenarios of implicit feedbacks. To remedy such gap, this paper proposes an implicit feedback-based group recommender system using probabilistic inference and non-cooperative game (GREPING) for IoT-based social media. Particularly, unknown process variables can be estimated from observable implicit feedbacks via Bayesian posterior probability inference. In addition, the globally optimal recommendation results can be calculated with the aid of non-cooperative game. Two groups of experiments are conducted to assess the GREPING from two aspects: efficiency and robustness. Experimental results show obvious promotion and considerable stability of the GREPING compared to baseline methods.

Due to the tremendous propagation loss of millimeter wave (mmWave) communication, device‐to‐device (D2D) communications are widely used over directional mmWave networks to improve network throughput. Most of existing works are mainly focused on concurrent transmission among D2D pairs that have no mutual interference. However, considering the propagation loss and directional interference features, the D2D pairs that interfere with each other can be used to improve network throughput. In this paper, we propose a time resource sharing scheme based on Stackelberg game for D2D pairs to further enhance network throughput. The interference D2D pairs can access to the time resource through paying the price for the interference caused by them. At the same time, the non‐interference D2D pairs can be scheduled. Concurrent transmission scheduling among interference D2D pairs is formulated as a non‐cooperative game, which enables a distributed transmission power control for interference D2D pairs. Moreover, the price strategy can be adjusted by setting interference threshold such that the transmission quality can be guaranteed. Simulation results show that the proposed scheme can achieve significant performance gains compared with classical concurrent transmission in terms of network throughput and energy efficiency. In addition, the convergence and utility of the game are also verified.