世界数字技术院（WDTA）：2024大语言模型安全性测试方法（中译版）（22页）.pdf

1、World Digital Technology Academy(WDTA)Large Language Model SecurityTesting MethodWorld Digital Technology Academy StandardWDTA AI-STR-02Edition:2024-04 WDTA 2024 All rights reserved.The World Digital Technology Standard WDTA AI-STR-02 is designated as a WDTAnorm.This document is the property of the

2、World Digital Technology Academy(WDTA)and isprotected by international copyright laws.Any use of this document,including reproduction,modification,distribution,or republication,without the prior written permission of WDTA,isprohibited.WDTA is not liable for any errors or omissions in this document.D

3、iscover more WDTA standard and related publications at https:/wdtacademy.org/.Version History*Standard IDVersionDateChangesWDTA AI-STR-021.02024-04Initial ReleaseForewordThe Large Language Model Security Testing Method,developed and issued by the World DigitalTechnology Academy(WDTA),represents a cr

4、ucial advancement in our ongoing commitment toensuring the responsible and secure use of artificial intelligence technologies.As AI systems,particularly large language models,continue to become increasingly integral to various aspects ofsociety,the need for a comprehensive standard to address their

5、security challenges becomesparamount.This standard,an integral part of WDTAs AI STR(Safety,Trust,Responsibility)program,is specifically designed to tackle the complexities inherent in large language models and providerigorous evaluation metrics and procedures to test their resilience against adversa

6、rial attacks.This standard document provides a framework for evaluating the resilience of large language models(LLMs)against adversarial attacks.The framework applies to the testing and validation of LLMsacross various attack classifications,including L1 Random,L2 Blind-Box,L3 Black-Box,and L4White-