Adversarial-Aware Deep Learning System Based on a Secondary Classical Machine Learning Verification Approach

. 2023 Jul 11;23(14):6287. doi: 10.3390/s23146287

Algorithm 1: Categorize Image Dataset.

[SETcrc, SETmis, SETadv] = Category (Image dataset, DNN classification results)
Input: ${x, r i g h t l a b e l} \in CIFAR - 100 (t e s t s e t), D N N m o d e l D N N (x), a d v_a t t a c k A M L (x)$
Output: The three categories of image dataset according to DNN model classification and AML results.
Initialize SETcrc, SETmis, SETadv to be all empty
for image $x \in$ CIFAR-100 do
if $D N N (x)$ is mispredict then
SETmis ← x
else
if $D N N (x)$ is correct and AML(x) fail then
SETcrc ← x
else
SETadv ← x
end if
end if
end for
return [SETcrc, SETmis, SETadv]