This paper introduces the concept of adversarial samples and explores the reasons for the generation of adversarial samples firstly.Secondly,it analyzes the classical adversarial samples attack methods from different fields and expounds the main adversarial samples defense methods from different research directions.Finally,it reviews the advantages and disadvantages of the existing research results and gives the future development trend of adversarial sample research.
关键词
深度学习/深度神经网络/对抗样本/对抗攻击与防御/人工智能
Key words
deep learning/deep neural network/adversarial samples/adversarial attacks and defen-ses/artificial intelligence
维普
万方数据