Group Seminar：L2范数、欧式空间、高斯分布和PCA；深度模型的攻击与防御

注: 本文是作者的自我总结,主要作为个人总结记录, 欢迎大家批评,交流. https://zhouxiaowei1120.github.io/#blogs

https://github.com/zhouxiaowei1120/UTSseminar/blob/master/Seminar 2019.2.9.md

Topic 1：L2 norm and Gaussian Distribution

1. norm and distance https://blog.csdn.net/gongem/article/details/51131397
2. Gaussian distribution https://www.jianshu.com/p/84afcb83a03c
3. PCA and data distribution （Gaussian） https://book.douban.com/annotation/22351713/
4. Gaussian and L2 distance https://blog.csdn.net/neu_chenguangq/article/details/79288905 https://discuss.gluon.ai/t/topic/1793/2 https://blog.csdn.net/qq_28038207/article/details/81006453
5. variance and distance https://www.cnblogs.com/nolonely/p/6477248.html

Topic 2：attack and defence

1. Adversarial Attacks and Defences: A Survey https://arxiv.org/abs/1810.00069

   • Evasion Attack:
   • Poisoning Attack：
   • Exploratory Attack：
   • Training Phase Capabilities.
     • Data Injection:
     • Data Modification
     • Logic Corruption
   • Testing Phase Capabilities：
     • White-Box Attacks
     • Black-Box Attacks
       • Non-Adaptive Black-Box Attack:
       • Adaptive Black-Box Attack:
       • Strict Black-Box Attack:

   Adversarial Goals：

   • Confidence Reduction
   • Misclassification
   • Targeted Misclassification
   • Source/Target Misclassification