Description of Invention
Adversarial Machine Learning (ML) is one of the biggest cyber threats to deep neural networks (DNN). Adversarial samples are crafted or created by an attacker to mislead a DNN model in making a decision. In order to ease the identification of adversarial attacks, a comparison study for visual analysis on adversarial images between t-SNE and UMAP are conducted. CNN is applied to extract the features before dimension compression through t-SNE and UMAP. The results show that t-SNE is more efficient in visualizing the adversarial attacks.