Clean label backdoor attack scenario #949
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
ebubae
reviewed
Nov 18, 2020
Comment on lines
+55
to
+71
| def poison_dataset(src_imgs, src_lbls, src, tgt, ds_size, attack, poisoned_indices): | ||
| # In this example, all images of "src" class have a trigger | ||
| # added and re-labeled as "tgt" class | ||
| poison_x = [] | ||
| poison_y = [] | ||
| for idx in range(ds_size): | ||
| if src_lbls[idx] == src and idx in poisoned_indices: | ||
| src_img = src_imgs[idx] | ||
| p_img, p_label = attack.poison(src_img, [tgt]) | ||
| poison_x.append(p_img) | ||
| poison_y.append(p_label) | ||
| else: | ||
| poison_x.append(src_imgs[idx]) | ||
| poison_y.append(src_lbls[idx]) | ||
| poison_x, poison_y = np.array(poison_x), np.array(poison_y) | ||
|
|
||
| return poison_x, poison_y |
There was a problem hiding this comment.
Take note that for this attack, we are only poisoning images in the target class. The attack.poison method takes the entire dataset and selects the poisoned data within the dataset, and returns the newly poisoned data points.
There was a problem hiding this comment.
Understood. The poison_dataset wrapper is only used with the backdoor object of type PoisoningAttackBackdoor, not the PoisoningAttackCleanLabelBackdoor object.
There was a problem hiding this comment.
I.e. it is used at eval time, not to poison the training data.
added 4 commits
November 18, 2020 18:54
ng390
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #950
Clean label backdoor attack scenario
Separate issue:
-Hyperparameter tuning so targeted attack success rate is higher.