多模态大语言模型应用于工业数据分类分级的初步研究与思考
An Initial Investigation on Applying Multi-modal Large Language Model on Industrial Data Classification and Grading
阮子禅 1包宏宇 1王文君1
作者信息
- 1. 上海观安信息技术股份有限公司,上海,200072
- 折叠
摘要
在万物互联互通的时代,工业领域数据安全有其独特风险:数据平台化汇聚、工业设备平台产品漏洞未修、工业数据暴露面增加、新兴技术带来新威胁.数据分类分级作为我国数字化进程的关键一环,是帮助企业在数据安全中排列风险优先级,在策略战略制定中提供重要依据,是从存储和检索角度提高运营效率、降低成本的重要前置任务.数据分类分级方法通常涉及大量人工干预和静态分类规则.这类方法不仅耗时、人工成本高,且难以有效处理庞大多样且异构的数据集.本文提出了结合多模态大语言模型的混合模型框架,是对新兴技术应用于工业数据分类分级的初步研究与思考,旨在推动工业数据分类分级领域方法创新.
Abstract
In today's interconnected world,data security in the industrial field poses unique risks:data aggregation to datacenters;unresolved vulnerabilities in equipment,platforms and products in industrial fields;increased exposure of industrial data due to internet;and new threats posed by emerging edging techniques.As a key part of China's digitalization process,data classification and grading is an important prerequisite task.Which helps enterprises prioritizing risks for their classified data,providing important inputs in enterprise-level,security-level and business-level decision-making processes,and reducing costs for data storage by eliminating redundant data.Traditional data classification and grading methods usually involve high level of manual intervention and large amounts of static rules setting,which is time-consuming,labor-intensive and error-prone.Additionally,it is difficult to effectively handle growing large,diverse,and heterogeneous data,which is often case in industrial enterprises.This paper proposes a hybrid neural network model framework that leverages multi-modal large language models(MLLM)to address the challenges in traditional industrial data classification and grading domain.This is a preliminary study and reflection on the application of MLLM in industrial data classification and grading task,aimed to promote technical and methodological innovations in the field of industrial data classification and grading.
关键词
工业领域/数据分类分级/多模态大语言模型/大模型Key words
Industrial Data/Data Classification and Grading/Multi-modal Large Language Model(MLLM)/Large Model(LM)引用本文复制引用
出版年
2024
NETL
NSTL
万方数据