GitHub - SyedIkram/Defence-Mechanisms-Against-One-Pixel-Attack: By understanding the One-Pixel attack and how it affects the neural network, it is clear that the location of the pixel plays a significant role for the image to be perturbed. Performing changes on the pixel colour and the position can help the neural network to classify the image accurately. Based on these, we have come up with approaches to make our network less sensitive to these kinds of attacks.

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
.ipynb_checkpoints		.ipynb_checkpoints
__pycache__		__pycache__
data		data
images		images
models		models
saved		saved
utils		utils
.DS_Store		.DS_Store
.gitattributes		.gitattributes
MAIN_1_Data_Augmentation.ipynb		MAIN_1_Data_Augmentation.ipynb
MAIN_2_Enhanced Resnet.ipynb		MAIN_2_Enhanced Resnet.ipynb
adv_img.png		adv_img.png
adv_img_423.png		adv_img_423.png
enhanced.py		enhanced.py
model.py		model.py
one_pixel.py		one_pixel.py
predictor.py		predictor.py
predictor_p.py		predictor_p.py
report.pdf		report.pdf
resnet56.th		resnet56.th
resnetmodel.py		resnetmodel.py
testing.ipynb		testing.ipynb
train.py		train.py

About

By understanding the One-Pixel attack and how it affects the neural network, it is clear that the location of the pixel plays a significant role for the image to be perturbed. Performing changes on the pixel colour and the position can help the neural network to classify the image accurately. Based on these, we have come up with approaches to ma…

attack pixel one defense de-noising