[1]田 鹏,左大义,高艳春,等.面向实际场景的人工智能脆弱性分析[J].计算机技术与发展,2021,31(11):129-135.[doi:10. 3969 / j. issn. 1673-629X. 2021. 11. 021]
　TIAN Peng,ZUO Da-yi,GAO Yan-chun,et al.Vulnerability Analysis of Artificial Intelligence in Real World[J].,2021,31(11):129-135.[doi:10. 3969 / j. issn. 1673-629X. 2021. 11. 021]
点击复制

面向实际场景的人工智能脆弱性分析()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
31
期数:
2021年11期
页码:
129-135
栏目:
网络与安全
出版日期:
2021-11-10

文章信息/Info

Title:
Vulnerability Analysis of Artificial Intelligence in Real World
文章编号:
1673-629X(2021)11-0129-07
作者:
田 鹏1; 2 ; 左大义1; 2 ; 高艳春1; 2 ; 陈海兵1; 2 ; 丁 灏1; 2
1. 中国电子科技集团第三十研究所,四川 成都 610000;
2. 中电科网络空间安全研究院有限公司,北京 100191
Author(s):
TIAN Peng1; 2 ; ZUO Da-yi1; 2 ; GAO Yan-chun1; 2 ; CHEN Hai-bing1; 2 ; DING Hao1; 2
1. China Electronics Technology Group Corporation 30,Chengdu 610000,China;
2. China Electronics Technology Research Institute of Cyberspace Security Co. ,Ltd. ,Beijing 100191,China
关键词:
人工智能安全; 安全威胁; 深度学习; 对抗样本; 对抗检测
Keywords:
artificial intelligence security; security thread; deep learning; adversarial example; adversarial detecting
分类号:
TP309
DOI:
10. 3969 / j. issn. 1673-629X. 2021. 11. 021
摘要:
人工智能技术广泛应用于自动驾驶、无人机、机器人等自主无人系统,是实现场景感知、情报获取、辅助决策等复杂功能的重要支撑。 因此, 研究人工智能技术的脆弱性和本身安全性问题引起了越来越多的关注。 对抗机器学习( adversarial machine learning)是机器学习和计算机安全领域的交叉学科,是人工智能算法普遍面临的挑战之一。 文中以实际场景下的人工智能安全性为出发点,梳理了对抗样本发展的起源,形成的机理以及发展脉络。 首先从攻击、防御两个方面探究各种方法的原理和优缺点;其次,在分析研究经典算法和适用场景的基础上,研究了在实际场景下智能技术面临的脆弱性和挑战;最后,针对在图像、语音、网络和软件应用等不同领域中所面临的挑战和未来发展趋势做了进一步的分析和展望。
Abstract:
Artificial intelligence is widely used in autonomous unmanned systems such as autonomous driving,unmanned aerial vehicles,robots,etc,which? ? is an important support to realize complex functions such as scene perception,intelligence acquisition,assistant decision-making and so on. Therefore, more and more attention has been paid for the vulnerability and security of artificial intelligence technology. Adversarial machine learning is an interdisciplinary subject in the field of machine learning and computer security. It is one of the challenges that artificial intelligence algorithms are facing. On the basis of the security of artificial intelligence in the actual scene,we introduce the origin,principle and development? ?of generating adversarial examples. Firstly,we explore the principles,advantages and disadvantages of attack and defense. Secondly,based on the analysis of classic algorithms and real world,the vulnerability and challenges faced in the actual scene are studied. Finally,we make a further study on the challenges and future development trend in different fields such as image,voice,network and software application.

相似文献/References:

[1]吴杨 矫文成 潘艳辉 李华.卫星网络加密算法安全性分析与攻击建模[J].计算机技术与发展,2011,(06):140.
　WU Yang,JIAO Wen-cheng,PAN Yan-hui,et al.Analysis of Cipher Security and Cipher Attack Modeling in Satellite Network[J].,2011,(11):140.
[2]王月,吕光宏,曹勇.软件定义网络安全研究[J].计算机技术与发展,2018,28(04):128.[doi:10.3969/ j. issn.1673-629X.2018.04.027]
　WANG Yue,LYU Guang-hong,CAO Yong.Research on Security of Software Defining Network[J].,2018,28(11):128.[doi:10.3969/ j. issn.1673-629X.2018.04.027]

更新日期/Last Update: 2021-11-10