MeanSparse: Post-Training Robustness Enhancement Through Mean-Centered Feature Sparsification
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866908582408093696 |
|---|---|
| author | Amini, Sajjad Teymoorianfard, Mohammadreza Ma, Shiqing Houmansadr, Amir |
| author_facet | Amini, Sajjad Teymoorianfard, Mohammadreza Ma, Shiqing Houmansadr, Amir |
| contents | We present a simple yet effective method to improve the robustness of both Convolutional and attention-based Neural Networks against adversarial examples by post-processing an adversarially trained model. Our technique, MeanSparse, cascades the activation functions of a trained model with novel operators that sparsify mean-centered feature vectors. This is equivalent to reducing feature variations around the mean, and we show that such reduced variations merely affect the model's utility, yet they strongly attenuate the adversarial perturbations and decrease the attacker's success rate. Our experiments show that, when applied to the top models in the RobustBench leaderboard, MeanSparse achieves a new robustness record of 75.28% (from 73.71%), 44.78% (from 42.67%) and 62.12% (from 59.56%) on CIFAR-10, CIFAR-100 and ImageNet, respectively, in terms of AutoAttack accuracy. Code is available at https://github.com/SPIN-UMass/MeanSparse |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_05927 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | MeanSparse: Post-Training Robustness Enhancement Through Mean-Centered Feature Sparsification Amini, Sajjad Teymoorianfard, Mohammadreza Ma, Shiqing Houmansadr, Amir Computer Vision and Pattern Recognition Cryptography and Security Machine Learning We present a simple yet effective method to improve the robustness of both Convolutional and attention-based Neural Networks against adversarial examples by post-processing an adversarially trained model. Our technique, MeanSparse, cascades the activation functions of a trained model with novel operators that sparsify mean-centered feature vectors. This is equivalent to reducing feature variations around the mean, and we show that such reduced variations merely affect the model's utility, yet they strongly attenuate the adversarial perturbations and decrease the attacker's success rate. Our experiments show that, when applied to the top models in the RobustBench leaderboard, MeanSparse achieves a new robustness record of 75.28% (from 73.71%), 44.78% (from 42.67%) and 62.12% (from 59.56%) on CIFAR-10, CIFAR-100 and ImageNet, respectively, in terms of AutoAttack accuracy. Code is available at https://github.com/SPIN-UMass/MeanSparse |
| title | MeanSparse: Post-Training Robustness Enhancement Through Mean-Centered Feature Sparsification |
| topic | Computer Vision and Pattern Recognition Cryptography and Security Machine Learning |
| url | https://arxiv.org/abs/2406.05927 |