Distribution-restrained Softmax Loss for the Model Robustness
This work addresses the robustness of deep learning models against attacks, which is an incremental improvement in loss function design for security applications.
The paper tackles the problem of improving deep learning model robustness by identifying that the distribution characteristics of softmax values for non-real label samples affect vulnerability to attacks, and proposes a loss function to suppress this diversity, resulting in enhanced robustness with minimal time overhead.
Recently, the robustness of deep learning models has received widespread attention, and various methods for improving model robustness have been proposed, including adversarial training, model architecture modification, design of loss functions, certified defenses, and so on. However, the principle of the robustness to attacks is still not fully understood, also the related research is still not sufficient. Here, we have identified a significant factor that affects the robustness of models: the distribution characteristics of softmax values for non-real label samples. We found that the results after an attack are highly correlated with the distribution characteristics, and thus we proposed a loss function to suppress the distribution diversity of softmax. A large number of experiments have shown that our method can improve robustness without significant time consumption.