Meta-learning aims to leverage prior knowledge from related tasks to enable a base learner to quickly adapt to new tasks with limited labeled samples. However, traditional meta-learning methods have limitations as they provide an optimal initialization for all new tasks, disregarding the inherent uncertainty induced by few-shot tasks and impeding task-specific self-adaptation initialization. In response to this challenge, this article proposes a novel probabilistic meta-learning approach called prototype Bayesian meta-learning (PBML). PBML focuses on meta-learning variational posteriors within a Bayesian framework, guided by prototype-conditioned prior information. Specifically, to capture model uncertainty, PBML treats both meta-and task-specific parameters as random variables and integrates their posterior estimates into hierarchical Bayesian modeling through variational inference (VI). During model inference, PBML employs Laplacian estimation to approximate the integral term over the likelihood loss, deriving a rigorous upper-bound for generalization errors. To enhance the model’s expressiveness and enable task-specific adaptive initialization, PBML proposes a data-driven approach to model the task-specific variational posteriors. This is achieved by designing a generative model structure that incorporates prototype-conditioned task-dependent priors into the random generation of task-specific variational posteriors. Additionally, by performing latent embedding optimization, PBML decouples the gradient-based meta-learning from the high-dimensional variational parameter space. Experimental results on benchmark datasets for few-shot image classification illustrate that PBML attains state-of-the-art or competitive performance when compared to other related works. Versatility studies demonstrate the adaptability and applicability of PBML in addressing diverse and challenging few-shot tasks. Furthermore, ablation studies validate the performance gains attributed to the inference and model components.

As an emerging technology, cloud-based Vehicular Ad-Hoc Network (VANET) can significantly improve the efficiency of transportation systems by alleviating traffic congestion and optimizing resource management. However, since the information transmitted in VANET is distributed in an open-access environment, security and privacy are now becoming vital concerns, especially when attackers own more and more resources. Recently, many tamper-proof devices (TPD)-based authentication schemes have been proposed to tackle the above-mentioned challenges, while most of them are dependent on ideal TPDs with very strong security assumptions. Moreover, when the aggregate signature verification fails, these schemes can do nothing but discard the aggregate signature completely, which makes them impractical in VANET. To address the above concerns, we propose a more realistic TPD-based authentication scheme with privacy-preserving features for cloud-based VANET. Specifically, the off-line self-updating method is utilized to update the data in TPD periodically to resist side-channel attacks. Furthermore, an efficient fine-grained error location algorithm is designed to quickly detect all invalid signatures within a problematic aggregate signature scenario. Our performance analysis shows that the proposed scheme outperforms the existing ones in terms of security, computation delay and communication overhead, thus more suitable in practical cloud-based VANET environment.

The Software Defined Network (SDN)-based space-ground integrated railway communication networks have attracted widespread attention from academia and industry. In such environments, the security of initial authentication and handover authentication for moving trains are two important challenges that need to be addressed. In this paper, a secure and efficient authentication key agreement scheme is proposed for the SDN-based space-ground integrated railway networks. Specifically, a lightweight mutual authentication mechanism based on the Number Theory Research Unit (NTRU) is proposed for the initial authentication process, which effectively prevents the unauthorized On-Board Unit (OBU) accessing networks. Then, according to the predictable path, we propose a key generation algorithm based on the hash chain and a fast key distribution mechanism based on the Chinese Remainder Theorem (CRT), which greatly reduce the calculation and communication burden of the key transmission process. On this basis, we adopt a hash-based message authentication code to achieve unified handover authentication in heterogeneous integrated railway networks. The Burrows-Abadi-Needham (BAN) logic proof and informal security analysis demonstrate that the proposed scheme can provide several robust security properties, including forward/backward security, universality, traceability, and resistance against quantum attacks. The performance evaluations show that our scheme outperforms other related schemes in computation cost, communication overhead, and performance under unknown attacks while guaranteeing higher security.

The energy management strategy (EMS) of plug-in fuel cell hybrid electric vehicles (P-FCHEV) is studied in this paper. Deep reinforcement learning (DRL) is a data-driven method that plays a crucial role in improving battery SoC maintenance, and fuel economy and extending fuel cell lifespan. However, existing methods struggle to balance optimal charge depletion and energy savings. This paper proposes a multi-constrained DRL-based EMS. Specifically, an energy management hierarchical framework is built by merging twin delayed deep deterministic policy gradient (TD3) with adaptive fuzzy control filtering. Then, a high-performance exploration technique is designed to accelerate the search for the optimal action, and a multi-objective adaptive penalty function based on the equivalent consumption minimization is constructed to balance fuel economy, battery power maintenance, and energy degradation. Finally, the charge depletion method is developed based on the SoC change prediction. Simulation results show that compared with the baseline TD3, the proposed EMS can reduce the fuel cell degradation rate by 10.49%, improve the SoC maintenance, and above 95% global optimum of the DP method. Furthermore, the average deviation from the terminal SoC is 1.92% under various scenarios, confirming that the proposed EMS can precisely achieve the optimal charge depletion.