SNN对抗攻击笔记

SNN对抗攻击笔记:

1. 解决SNN对抗攻击中脉冲与梯度数据格式不兼容性以及梯度消失问题:

  • G2S Converter、Gradient Trigger[1]

 

2. 基于梯度的对抗攻击方式:

  • FGSM、BIM[1]

 

3. 采用的神经元模型:

  • 迭代LIF神经元模型[1]

 

4. 图像转化到脉冲序列的采样方式:

  • Bernoulli采样[1]
  • Poisson编码器[2]

 

5. 替代梯度法:

  • 阶跃函数[1]

 

6. 影响对抗攻击效果的因素分析:

  • 损失函数和发放阈值(倒数第二层)[1]

 

7.对抗攻击类型:

  • 白盒攻击、目标/非目标攻击[1]

 

Reference:

[1] Liang L , Hu X , Deng L , et al. Exploring Adversarial Attack in Spiking Neural Networks with Spike-Compatible Gradient[J]. 2020.

[2] Sharmin S , Rathi N , Panda P , et al. Inherent Adversarial Robustness of Deep Spiking Neural Networks: Effects of Discrete Input Encoding and Non-Linear Activations[J]. 2020.

[3] Sharmin S , Panda P , Sarwar S S , et al. A Comprehensive Analysis on Adversarial Robustness of Spiking Neural Networks[J]. 2019.

上一篇:Towards Evaluating the Robustness of Neural Networks(翻译,侵删)


下一篇:[Machine Learning] Neural Networks: Representation Quiz