NERO-Net: A Neuroevolutionary Approach for the Design of Adversarially Robust CNNs
This addresses the need for robust models in safety-critical applications by focusing on architectural design rather than training methods, though it is incremental as it builds on neuroevolution and robust training concepts.
The paper tackled the problem of designing convolutional neural networks with intrinsic adversarial robustness by proposing NERO-Net, a neuroevolutionary approach that evolved architectures achieving 47% adversarial accuracy against FGSM and 93% clean accuracy on CIFAR-10 without adversarial training during evolution.
Neuroevolution automates the complex task of neural network design but often ignores the inherent adversarial fragility of evolved models which is a barrier to adoption in safety-critical scenarios. While robust training methods have received significant attention, the design of architectures exhibiting intrinsic robustness remains largely unexplored. In this paper, we propose NERO-Net, a neuroevolutionary approach to design convolutional neural networks better equipped to resist adversarial attacks. Our search strategy isolates architectural influence on robustness by avoiding adversarial training during the evolutionary loop. As such, our fitness function promotes candidates that, even trained with standard (non-robust) methods, achieve high post-attack accuracy without sacrificing the accuracy on clean samples. We assess NERO-Net on CIFAR-10 with a specific focus on $L_\infty$-robustness. In particular, the fittest individual emerged from evolutionary search with 33% accuracy against FGSM, used as an efficient estimator for robustness during the search phase, while maintaining 87% clean accuracy. Further standard training of this individual boosted these metrics to 47% adversarial and 93% clean accuracy, suggesting inherent architectural robustness. Adversarial training brings the overall accuracy of the model up to 40% against AutoAttack.