The intellectual property (IP) rights of artificial intelligence (AI) models are the prerequisite for the flourishing machine learning as a service (MLaaS) market. Recently, studies on AI model copyright protection have been burgeoning, but still face challenges. On the one hand, current AI model copyright sharing is unfair. Collaborative MLaaS nodes that contribute more training resources usually cannot gain proportional copyright benefits, severely dampening their enthusiasm for participating in AI model trading. On the other hand, untrustworthy third-party verifiers may act erroneously or unreliably when handling AI model ownership disputes, which greatly diminishes the credibility of the verification results. To address the above challenges, we propose AIIP-Chain, a fair AI model copyright sharing with credible ownership verification framework. First, we design a contribution-aware training effort evaluation scheme to ensure the fairness of copyright sharing, in which MLaaS nodes obtain copyright shares according to their consumed training resources. Moreover, to enhance the trustworthiness of the verification results, we propose a crowdsourcing credible AI model ownership verification scheme that balances node motivation, verification cost, and node reputation. Finally, we evaluate the fairness and credibility of AIIP-Chain and the results demonstrate the effectiveness of our schemes.

Collaborative perception (CP) is a critical enabler for enhancing situational awareness, traffic safety, and mobility in connected autonomous vehicles (CAVs). By integrating sensory data from multiple CAVs, CP effectively mitigates perceptual blind spots, reduces the likelihood of traffic accidents, and alleviates congestion within complex environments. To advance CP capabilities within dynamic and resource-constrained network conditions, this paper proposes a proactive collaborative perception strategy that enables CAVs to selectively share perceptual data with other CAVs based on real-time network status and anticipated perceptual demands. Specifically, a collaborative framework, integrating communication and perception, is designed to enhance CP performance with data processing and fusion techniques. Within this framework, an innovative adaptive data compression algorithm is introduced, which dynamically adjusts the compression ratio based on the monitored real-time signal-to-noise ratio, therefore optimizing the data transmission efficiency. Additionally, a mathematical model is formulated to jointly optimize the communication and perception resources, and a multi-agent reinforcement learning algorithm based on Global State Proximal Policy Optimization (GSPPO) is developed to enhance communication resources distribution and CAV selection in complex and dynamic network environments. Experimental results demonstrate that the proposed proactive CP strategy can effectively reducing communication latency without compromising perception accuracy.

Natural disasters such as earthquakes can cause damage to critical infrastructures and limit access to vital information, making it difficult for disaster response teams to respond effectively. Unmanned aerial vehicles (UAVs) have the potential to aid and provide real-time information for disaster response teams, however, the need to process distributed learning for huge amounts of interconnected nodes in a graph network poses several challenges. First, distributed learning in graph networks for UAVs is still an open issue, making it difficult to train and share models on such networks. Second, such a network can leak privacy-sensitive information, making it harder to ensure data security. To address these challenges, we propose, in this paper, a novel privacy and blockchain-empowered UAVs-enabled decentralized graph federated learning (DPBE-DGFL) framework for disaster response. The framework includes three phases: (i) local model training utilizing stochastic gradient descent with differential privacy, (ii) model weights integrity authentication using blockchain to ensure secure and efficient sharing of model weights, and (iii) final validator selection and model weights aggregation using a dedicated proof-of-stake, (DPoS), consensus mechanism to ensure efficient and decentralized consensus while maintaining security and integrity. Our DPBE-DGFL framework was evaluated using extensive simulations on EMNIST and real-world disaster datasets from Tonga. The results show that it offers a promising solution for privacy-preserving federated learning in graph networks, balancing privacy protection and model accuracy while maintaining latency, communication, and computational efficiency.

With the proliferation of connected and automated vehicles (CAVs), densely distributed edge computing nodes have emerged on roadways. Consequently, leveraging CAVs as a mobile computing platform can integrate idle vehicle resources to provide computational services for ubiquitous Internet of Things (IoT) devices. Numerous studies have investigated task offloading strategy in the systems with full CAVs penetration. It is expected that the coexistence of CAVs and human-driven vehicles (HDVs) in mixed traffic systems will continue for a considerable period. However, due to the impact of HDVs on communication performance, the task offloading model designed for the systems with full CAVs penetration are no longer applicable in mixed traffic systems. We explore task offloading schemes using CAVs as a mobile computing platform in mixed traffic systems to address this issue. Specifically, we first model the communication model in mixed traffic systems, taking into account the influence of HDVs on link interference, the alteration of path loss due to the impact of HDVs on routing, and the additional sensing tasks arising from the inability of HDVs and CAVs to communicate. Subsequently, considering that delay and energy consumption are crucial factors affecting the performance of CAVs as a mobile computing platform, we formulate the task offloading scheme as an optimization problem. Additionally, we employ a distributed offloading based on deep learning (DODL) algorithm to obtain approximately optimal offloading decisions. Simulation results demonstrate the effectiveness of the proposed model in mixed traffic systems. By employing the DODL algorithm, the CAVs as a mobile computing platform can achieve enhanced performance in terms of convergence, thereby advancing the development of autonomous driving in mixed traffic systems.

Unmanned Aerial Vehicles (UAVs) have attracted extensive attention from both industry and academia owing to their high mobility, line-of-sight (LoS) characteristics of air-toground (A2G) channels, and low cost. However, the broadcast nature of wireless transmission and the LoS characteristics of A2G channels are vulnerable to eavesdropping attack, which leads to severe security issues. To enhance the security of UAV communication, we propose a framework that multiple UAVs cooperate to resist attacks (MURA). Specifically, we first propose an efficient incentive scheme based on the coalitional game to encourage UAVs to join the coalition. We prove that each UAV can maximize its utility by joining the coalition to form a grand coalition. Then, a secure UAV communication scheme is proposed to resist eavesdropping attack. Two types of scenarios are considered for UAV communication. In a completely rational scenario, in which participants make decisions aiming to maximize their utility, we utilize the Stackelberg game to model the interactions between UAVs and attacker. The existence and uniqueness of the equilibrium solution are proved, and the equilibrium solution is obtained. In an imperfectly rational scenario, the prospect theory (PT) is applied to capture the underlying rationality of the players. The PT valuations of the players, i.e., UAV and attacker, are deduced in detail. Meanwhile, the convergence of the PT valuations of UAV and attacker is proved. Finally, extensive simulation results show that the proposed scheme can effectively improve the utility of legal UAVs and ensure the security of the UAV networks compared with benchmarks.

To investigate the diversified applications in vehicular networks, artificial intelligence, intelligent edge computing, and vehicular networks are combined. By offloading computation tasks to devices close to vehicles, Vehicular Edge Computing (VEC) has emerged as a new computing paradigm to tackle the problem. Most existing VEC methods simply slice the application into subtasks for offloading purposes without considering the dependencies between subtasks. In practice, the dependency information is critical to the efficiency of offloading strategies. If a subtask requires the computation result of another subtask, the latter has to be processed before the former is finished. In this paper, we propose a deep reinforcement learning based offloading strategy for multi-vehicle collaboration VEC, with task dependency taken into account. With the proposed strategy, we formulate the offloading problem as an Markov Decision Process (MDP) and use the Sequence-to-Sequence (S2S) neural network to represent the policy/value function of the MDP. Furthermore, we train the S2S neural network to obtain the appropriate offloading policy using the Proximal Policy Optimization (PPO) technique. Our simulation results indicate that, by considering task dependencies during offloading, the proposed strategy outperforms existing methods in effectively reducing task offloading latencies.

Unmanned aerial vehicles (UAVs) have attracted extensive attention from both industry and academia owing to their high mobility, and characteristics of line of sight (LoS) propagation. However, wireless communication is vulnerable to eavesdropping attacks because of the broadcast characteristics. To enhance secure UAV communications with the ground nodes, we propose a novel framework that multiple UAVs cooperate to resist attack (MURA). First, we propose an incentive mechanism based on coalitional game to encourage legal UAVs to join the coalition. We prove that each legal UAV can only maximize its profits by joining the coalition to form a major coalition. Then, a secure UAV communication scheme is proposed to resist the eavesdropping attacks. Two types of scenarios are considered for the UAV communication: in a completely rational scenario, we utilize the Stackelberg game to model the interactions between the legal UAVs and attacker. In an imperfectly rational scenario, the cumulative prospect theory (PT) is applied to the game to capture the underlying rationality of the players. Finally, simulation results show that the proposed scheme can significantly improve the security of the UAV network compared with traditional schemes.

The Fintech index has been more active in the stock market with the Fintech industry expanding. The prediction of the Fintech index is significant as it is capable of instructing investors to avoid risks and provide guidance for financial regulators. Traditional prediction methods adopt the deep neural network (DNN) or the combination of genetic algorithm (GA) and DNN mostly. However, heavy computational load is required by these algorithms. In this paper, we propose an integrated artificial intelligence‐based algorithm, consisting of the random frog algorithm (RF), GA, and DNN, to predict the Fintech index. The proposed RF‐GA‐DNN prediction algorithm filters the key input variables and optimizes the hyperparameters of DNN. We compare the proposed RF‐GA‐DNN with the traditional GA‐DNN in terms of convergence time and prediction accuracy. Results show that the convergence time of GA‐DNN is up to 20 hours and its prediction accuracy is 97.4%. In comparison, the convergence time of our RF‐GA‐DNN is only about 1.5 hours and the prediction accuracy reaches 97.0%. These results demonstrate that the proposed RF‐GA‐DNN prediction algorithm significantly reduces the convergence time with the promise of competitive prediction accuracy. Thus, the proposed algorithm deserves to be widely recommended for predicting the Fintech index.