摘要
作为一项能够融入社会生活各个方面的技术,人工智能正在引发一系列关键变革,同时也带来了在政治、军事、社会治理等各个方面的安全威胁,既需要制定人工智能的安全框架保障人工智能系统本身不被对手利用,又需要发展可信任的人工智能机制提升人工智能系统本身的安全性。本文首先从定义、分类和度量的角度解析了人工智能的内涵,列举了人工智能可能带来的安全威胁。在威胁分析的基础上,参考传统的安全模型提出了以安全策略和安全边界为核心的人工智能安全框架。最后,结合人类社会中的管理制度和传统信息系统的安全管理规则提出人工智能系统需要具备容错机制、权限机制、可解释机制等特点,发展人工智能需要将可信任放在首位。
As a technology that can be integrated into all aspects of social life,Al is triggering a series of key changes,but also brings security threats in political,military,social governance and other aspects.It is necessary to design a security framework of Al to ensure that other opponents cannot take advantage of,and it is also important to develop trusted Al mechanisms to enhance the security of Al system itself.This paper firstly analyzes Al from the point of definition,classification and measurement,and lists the security threats that Al may bring.On the basis of threat analysis and traditional security model,an artificial intelligence security framework based on security strategy and security boundary is proposed.Finally,combined with the management system in human society and the security management rules of traditional information systems,it is pointed out that dependable Al systems need fault-tolerant mechanism,privilege mechanism and interpretable mechanism,and the further development of Al needs to put trustworthiness first.
作者
张林超
张欣海
ZHANG Lin-chao;ZHANG Xin-hai(China Academy of Electronics and Information Technology,Beijing 100041,China)
出处
《中国电子科学研究院学报》
北大核心
2019年第12期1253-1258,共6页
Journal of China Academy of Electronics and Information Technology
基金
国家重点研发计划(2017YFC0820503)
中国电科装备预研联合基金(6141BO8O8O1O1,6141B08010102)。
关键词
人工智能
安全框架
可信任系统
artificial intelligence
security architecture
trusted system
作者简介
张林超(1987-),男,河南人,高级工程师,主要研究方向为网络协议、人工智能及大数据应用技术,E-mail:hune213@163.com;张欣海(1975-),男,辽宁人,研究员级高级工程师,主要研究方向为大数据与人工智能应用技术。
