Traditional autonomous driving systems often struggle to connect high-level reasoning with low-level control, leading to suboptimal and sometimes unsafe behaviors. Recent advances in multimodal large language models (MLLMs), which process both visual and textual data, offer an opportunity to unify perception and reasoning. However, effectively embedding precise safety knowledge into MLLMs for autonomous driving remains a significant challenge. To address this, we propose SafeAuto, a framework that enhances MLLM-based autonomous driving by incorporating both unstructured and structured knowledge. First, we introduce a Position-Dependent Cross-Entropy (PDCE) loss to improve low-level control signal predictions when values are represented as text. Second, to explicitly integrate safety knowledge, we develop a reasoning component that translates traffic rules into first-order logic (e.g., "red light =$>$ stop") and embeds them into a probabilistic graphical model (e.g., Markov Logic Network) to verify predicted actions using recognized environmental attributes. Additionally, our Multimodal Retrieval-Augmented Generation (RAG) model leverages video, control signals, and environmental attributes to learn from past driving experiences. Integrating PDCE, MLN, and Multimodal RAG, SafeAuto outperforms existing baselines across multiple datasets, enabling more accurate, reliable, and safer autonomous driving. The code is available at https://github.com/AI-secure/SafeAuto.

Cloud control systems (CCSs) are evolving rapidly, requiring secure communication channels to protect critical remote control tasks. This article introduces an efficient encryption mechanism configured for securing the physical layer communication in these systems while transmitting data over optical fiber networks using orthogonal frequency division multiplexing active optical networks. The proposed scheme takes advantage of the hypersensitive chaotic properties of the Lorenz map to provide robust multidimensional encryption and confidentiality. The encryption process involves arranging bulk encoded codewords in a table structure shape, with each column representing individual codewords encoded by pipelined successive cancellation polar encoding. A Lorenz map is then utilized to generate three distinct chaotic keys, which are employed to reindex the rows and columns of the table structured codeword along with the subcarrier remapping in the constellation map. This reindexing operation provides an additional layer of security against potential brute-force attacks. Performance metrics such as computational efficiency, security robustness, and resistance to various noise and distortion sources are being evaluated. The experimental results demonstrate promising levels of security and resilience against potential threats on the physical layer of communications. The scheme's compatibility with the technical environment ensures seamless integration into existing CCSs infrastructure, making it a resilient solution for securing critical physical layer transmissions.

An appropriate selection is of great importance to Web service composition. Current approaches normally focus on the average performance of quality of service (QoS) of Web services while not paying more attention to the performance such as stability considering the changing environments. In addition, Web services are differentiated from others with same functions which are provided by various physical origins of service requesters and service providers, which will lead to unexpected QoS of Web service system. To solve the two issues, this paper presents a QoS-aware strategy for Web service composition based on cloud model, and correspondingly designs a multi-objective Web service composition algorithm based on quantum genetic algorithm. The simulation based on a typical dataset validates the proposed approach finally.