Intriguing properties of neural networks, Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian Goodfellow, Rob Fergus, 2013arXiv preprint arXiv:1312.6199DOI: 10.48550/arXiv.1312.6199 - This is the seminal paper that introduced the concept of adversarial examples, demonstrating that small, imperceptible perturbations can cause deep neural networks to misclassify inputs.
Explaining and Harnessing Adversarial Examples, Ian J. Goodfellow, Jonathon Shlens, Christian Szegedy, 2014arXiv preprint arXiv:1412.6572DOI: 10.48550/arXiv.1412.6572 - This paper explains the cause of adversarial examples, links them to the linear nature of deep neural networks, and introduces the Fast Gradient Sign Method (FGSM) for generating them.
Adversarial Machine Learning: A Survey of Recent Advances, Xiaoyong Yuan, Pan He, Qiming Zhu, Xin Li, Shouhuai Xu, 2019ACM Computing Surveys (CSUR), Vol. 52 (Association for Computing Machinery (ACM))DOI: 10.1145/3313936 - This comprehensive survey provides a broad overview of adversarial machine learning, including formal definitions of adversarial examples, various attack methods (like those using Lp norms), and defense strategies.