信息通信技术与政策

信息通信技术与政策

信息通信技术与政策 ›› 2021, Vol. 47 ›› Issue (8): 56-63.

上一篇    下一篇

人工智能系统安全综述*

An overview of artificial intelligence system security

陈磊1,2,李雅静3
  

  1. 1.国防科技大学,长沙410073;
    2.中国工程院战略咨询中心,北京100088;
    3.麒麟软件有限公司,北京100190
  • 出版日期:2021-08-15 发布日期:2021-08-29
  • 作者简介:
    陈磊摇摇国防科技大学博士研究生,中国工程院战略咨询中心助理研究员,研究方向为科技政策、人工智能安全等
    李雅静摇通信作者。麒麟软件有限公司,研究方向为科技政策、新媒体等
  • 基金资助:
    中国工程院2019 年重大咨询研究项目(No.2019-ZD-1)资助

CHEN Lei1,2, LI Yajing3,#br#   

  1. 1.National University of Defense Technology, Changsha 410073, China;
    2.Center for Strategic Studies, Chinese Academy of Engineering, Beijing 100088, China;
    3.KylinSoft Beijing 100190, China
  • Online:2021-08-15 Published:2021-08-29

摘要: 人工智能的迅速发展使得人们越来越关注人工智能技术对社会的潜在影响。对人工智能安全技术的基本组成,及内生安全、衍生内生安全进行了研究。从技术上,分析了机器学习由于奖励函数、数据变化以及算法探索等方面的安全性挑战;从应用上,提出了伦理和法制等方面值得研究和解决的问题;从国家安全角度,提出了技术的自主可控性;最后,展望了人工智能安全未来的研究方向。

关键词: 人工智能, 安全, 机器学习, 深度学习

Abstract: The rapid development of artificial intelligence makes people pay more and more attention to the potential impact of artificial intelligence technology on society.In this paper, the basic components of artificial intelligence safety technology are introduced, and the research status at home and abroad is expounded from the research direction of endogenous safety and derivative endogenous safety.In terms of technology, the security challenges of machine learning due to reward function, data change and algorithm exploration are analyzed.In terms of application, the ethical and legal issues worth studying and solving are proposed.From the perspective of national security, the autonomous controllability of the technology is proposed.Finally, the future research direction of artificial intelligence security is prospected.

Key words: artificial intelligence, safety, machine learning, deep learning