【总结】Obfuscated Gradients give a false sense of security:circumventing defenses to adversaria..

《Obfuscated Gradients give a false sense of security:circumventing defenses to adversarial examples》

相关文献

1、神经网络对于对抗样本的敏感性:
Szegedy et al. 2013
Biggio et al. 2013
2、iterative optimization-based attacks
Kurakin et al. 2016a
Madry et al. 2018
Carlini & Wagner 2017c
3、gradient masking
Papernot et al. 2017
4、Expectation Over Transformation
Athalye et al. 2017
5、ResNet
Zagoruyko & Komodakis, 2016
He et al. 2016
6、backpropagation
Rumelhart et al. 1986
7、降噪去对抗干扰
Guo et al. 2018
避开这种防御:Carlini & Wagner, 2017b
8、

三种obfuscated gradients:

shattered gradients:破碎梯度,产生于有意的不可导操作或者无意间数值不稳定造成
stochastic gradients:随机梯度取决于测试时间的随机性
vanishing/exploding gradients:梯度消失、梯度爆炸,很深的计算中产生的无用梯度

Attack Techniques

1、Backward Pass Differentiable Approximation (BPDA)
因为 g ( x ) x ,
所以 x g ( x ) x x = 1
所以 x f ( g ( x ) ) | x = x ^ x f ( x ) | x = g ( x ^ )

BPDA:
g ( x ) f i ( x ) 只在backward pass上用 g ( x ) 代替 f i ( x ) , 发现比forward和packward都替换的效果好得多
可以用来解决shattered gradients

2、Expectation over Transformation
用Expectation over Transformation来正确计算输入所期望的转变的梯度,来对抗让输入随机转变的防御

3、Reparameterization
solve vanishing/exploding gradients
make x = h ( z ) 使得 g ( h ( z ) ) = h ( z ) h ( · ) 是可微的,这样,可以通过 f ( h ( z ) ) 来计算梯度

Case study

1、Non-obfuscated Gradients
(1) adversarial training
用对抗样本训练,直到能正确分类
给定训练数据 X 和损失函数 l ( · ) , 一般的训练是选择参数

θ = a r g θ m i n E x X l ( x ; F θ )

而对抗训练解决
θ = a r g θ m i n E x X [ m a x δ [ ϵ , ϵ ] N   l ( x + δ ; F θ ) ]

(不太懂)

猜你喜欢

转载自blog.csdn.net/qq_37903108/article/details/81001924