计算机科学
伪装
对手
深度学习
人工智能
Guard(计算机科学)
计算机安全
信息安全
计算机视觉
程序设计语言
作者
Song Yuchen,Dejin Tang,Xiaoming Zhou,Yuanchen Song
标识
DOI:10.1145/3377713.3377787
摘要
At present, with the breakthrough and application of in-depth learning technology in the field of artificial intelligence, the content of information acquired through interpretation of remote sensing images is more and more abundant, and how to guard information security has become a new technology hotspot. [1] In order to camouflage and conceal our important objects, it is necessary to attack the acquired remote sensing images to remove and confuse the target information and mislead the "enemy" to make wrong image analysis, which is a means to protect information security.[2] In this paper, the mainstream algorithm principle is introduced based on machine deep learning network technology. Aiming at remote sensing image data poisoning attack and sample attack in the process of deep learning network training, the purpose of tampering with original image data and hiding characteristic targets is realized.
科研通智能强力驱动
Strongly Powered by AbleSci AI