中圖分類號(hào)：TP393.08 文獻(xiàn)標(biāo)志碼：A DOI: 10.16157/j.issn.0258-7998.256839
中文引用格式： 蔡超雄. 大模型網(wǎng)絡(luò)安全風(fēng)險(xiǎn)識(shí)別與評(píng)價(jià)體系研究[J]. 電子技術(shù)應(yīng)用，2026，52(2)：45-51.
英文引用格式： Cai Chaoxiong. Research on the identification and evaluation system of network security risks in large language models[J]. Application of Electronic Technique，2026，52(2)：45-51.

Research on the identification and evaluation system of network security risks in large language models

Abstract： This study aims to comprehensively identify the network security risks of multimodal large language models, construct a security evaluation system, and propose response strategies to enhance their security and reliability. The research method first analyzes the risk categories, constructs a risk identification model through multimodal data fusion, and uses the Analytic Hierarchy Process and Fuzzy Comprehensive Evaluation Method to construct a security evaluation system. Then, the effectiveness and stability of the model and system are verified through experiments. The research results indicate that the risk identification model can accurately identify risks, and the security evaluation system can reasonably assess the security situation. This study provides innovative ideas for the security risk identification of multimodal large language models. The constructed model and system can significantly enhance the risk identification ability, providing theoretical support for the security research of multimodal large language models and promoting the safe development of large language models.

Key words : multimodal large language model；network security； risk identification；analytic hierarchy process；fuzzy comprehensive evaluation method；security evaluation system；response strategy