Adversarial Machine Learning: Attacks, Defenses, and the Path Towards Trustworthy AI

Abstract

Deep neural networks achieve strong performance on perception tasks but remain vulnerable to adversarial examples—imperceptibly perturbed inputs that induce confident misclassification. This dissertation reviews the adversarial attack–defence landscape and reports CIFAR-10 experiments using ResNet-18. It compares a standard baseline, a PGDadversarially trained model, and a model obtained from RobustBench under FGSM, PGD-20, and AutoAttack.

Categories

Framework mappings

Suggested from the entry's categories.

@article{pang2026adversarial,
  title = {{Adversarial Machine Learning: Attacks, Defenses, and the Path Towards Trustworthy AI}},
  author = {Jieyao Pang},
  year = {2026},
  month = jul,
  journal = {International Journal of Innovative Science and Research Technology},
  doi = {10.38124/ijisrt/26jun1631},
  url = {https://doi.org/10.38124/ijisrt/26jun1631}
}