6 August 2024 in IEEE Transactions on Intelligent Transportation Systems
Due to the increasing number of electric vehicles and limited charging stations, charging service providers need to consider the minimum range of electric vehicles and the congestion level of charging stations to formulate pricing strategies and guide charging vehicles. However, this service faces the challenges of privacy protection and utility coupling between CSs and EVs. Besides, the minimum driving range is also affected by the time-varying data characteristics of vehicles. Therefore, this paper proposes an FL-assisted stable charging matching strategy considering congestion factors and range constraints. Specifically, we employ a federated learning scheme to predict the congestion factor of each charging station by learning the complex hidden characteristics of vehicles under the condition of privacy protection. Then, the pricing strategy and utility function are updated based on the congestion factor. Furthermore, under the driving range constraint, we transform the impact of the congestion factor on the expected utility into a regret value and use the Gale-Shapley algorithm based on regret minimization to achieve stable matching between vehicles and charging stations. Experimental results demonstrate the scheme’s effectiveness and that the charging success rate is improved by up to 11.7% compared with mainstream methods.
16 March 2023 in Transactions on Emerging Telecommunications Technologies
28 December 2023 in Knowledge-Based Systems
16 June 2022 in Security and Communication Networks
Named data networking (NDN) is a promising alternative data dissemination technology of TCP/IP communication networks, which can bring out much more cost-effective and resilient communication in a highly mobile environment. However, due to the feature of NDN, content poisoning comes out as a potential threat. Hence, state-of-the-art studies introduce network layer approaches based on name-key binding, in which the producer notifies routers of the bindings of names and key values. Key values include publisher public key digest or content digest. Routers check key values to determine whether incoming data packets have been poisoned. Unfortunately, the approaches lead to more vulnerabilities in dynamic content poisoning because attackers can impersonate the producer to alter or fabricate the bindings. Thus, we introduce a consumer-oriented two-phased lightweight security scheme, which consists of an end-to-end authentication and a packet-level name-key query mechanism. Specifically, the name-key bindings are authenticated via an additional verification by the consumer. Furthermore, we also introduce a novel trust model to help routers to determine and disconnect from the malicious nodes. Finally, our extensive experimental results demonstrate that the scheme can work effectively in improving the vulnerability of existing studies on dynamic content poisoning and lowering the system overhead simultaneously.
12 July 2024 in IEEE Internet of Things Journal
Federated learning coordinates distributed data sets to train models, which brings the significant impact of data selection on model performance. Personalized differential privacy, however, introduces heterogeneity into the vehicular data sets: the higher privacy protection may reduce the contribution of local models to model convergence. Therefore, the goal of this article is to dynamically optimize the combination of data sets to tackle the heterogeneity in differential private federated learning in Internet of Vehicles. This is extremely challenging without direct data access and a visible training process. Therefore, we propose an efficient hierarchical data selection method. First, the utility is evaluated using the convergence bound derived from the noise function and the cost function. Accordingly, a collection of high-value clients is selected to maximize the potential contribution of the combination to the global model. Then, we design an optimization function based on the unknown variables within the convergence bound and develop a low-complexity algorithm to approximate the sampling probability. Meanwhile, the aggregation weight of each model is adjusted to ensure unbiased estimation. Experimental results on two real-world trajectory data sets show that the scheme can reduce the meter error by 8.90% and 15.97%, respectively, and improve the convergence speed by 23.9% and 27.1%, respectively.