图 书 馆
对机器学习系统的对抗性攻击Adversarial Attacks On ML Systems |
|
| 课程网址: | http://videolectures.net/textSpeechDialogue_raj_adversarial_attac... |
| 主讲教师: | Bhiksha Raj |
| 开课单位: | 卡内基梅隆大学 |
| 开课时间: | 2019-10-08 |
| 课程语种: | 英语 |
| 中文简介: |  随着神经网络分类器在从语音识别和图像分类到各种自然语言处理任务甚至识别恶意软件的各种任务中越来越成功，第二个有点令人不安的发现也出现了。可以通过精心设计的输入来欺骗这些系统，这些输入在外行观察者看来是自然数据，但会导致神经网络以随机甚至有针对性的方式错误分类。 在本次演讲中，我们将讨论原因这种攻击是可能的，以及设计、识别和避免这种精心设计的“对抗性”输入的攻击的问题。 |
| 课程简介: | As neural network classifiers become increasingly successful at various tasks ranging from speech recognition and image classification to various natural language processing tasks and even recognizing malware, a second, somewhat disturbing discovery has also been made. It is possible to fool these systems with carefully crafted inputs that appear to the lay observer to be natural data, but cause the neural network to misclassify in random or even targeted ways. In this talk we will discuss why such attacks are possible, and the problem of designing, identifying, and avoiding attacks by such crafted "adversarial" inputs. |
| 关 键 词: | 神经网络分类器; 自然数据; 机器学习系统 |
| 课程来源: | 视频讲座网 |
| 数据采集: | 2021-06-18：yumf |
| 最后编审: | 2021-06-18：yumf |
| 阅读次数: | 40 |