defense arXiv Oct 10, 2025 · Oct 2025
Soroush Mahdi, Maryam Amirmazlaghani, Saeed Saravani et al. · Amirkabir University of Technology · Sharif University of Technology
Adversarial training defense that recycles past-epoch adversarial examples to improve accuracy-robustness trade-off without external data
Input Manipulation Attack vision
In this paper, we propose a new approach called MemLoss to improve the adversarial training of machine learning models. MemLoss leverages previously generated adversarial examples, referred to as 'Memory Adversarial Examples,' to enhance model robustness and accuracy without compromising performance on clean data. By using these examples across training epochs, MemLoss provides a balanced improvement in both natural accuracy and adversarial robustness. Experimental results on multiple datasets, including CIFAR-10, demonstrate that our method achieves better accuracy compared to existing adversarial training methods while maintaining strong robustness against attacks.
cnn Amirkabir University of Technology · Sharif University of Technology
ML Security Papers