I have read the use of Targeted Adversarial Attacks for making the model perform better. But can we change the bias of the neural networks and control the outcome of the network rather than changing the input. if yes, can you share some resources or research papers on targeted bias in neural networks?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
No, the bias is part of the network, not one of the inputs, so an adversary has no ability to manipulate it.
Unless the adversary can change the network, in which case they don't need to use any tricks, because they can just change it to one that outputs what they want!
