Improving Machine Learning Robustness via Adversarial Training

no code implementations • 22 Sep 2023 • Long Dang, Thushari Hapuarachchi, Kaiqi Xiong, Jing Lin

Moreover, in the non-IID data case, the natural accuracy drops from 66. 23% to 57. 82%, and the robust accuracy decreases by 25% and 23. 4% in C&W and Projected Gradient Descent (PGD) attacks, compared to the IID data case, respectively.

Federated Learning

ML Attack Models: Adversarial Attacks and Data Poisoning Attacks

no code implementations • 6 Dec 2021 • Jing Lin, Long Dang, Mohamed Rahouti, Kaiqi Xiong

Many state-of-the-art ML models have outperformed humans in various tasks such as image classification.

Data Poisoning Image Classification