Adversarial Robustness in Multi-Task Learning: Promises and Illusions
This addresses the vulnerability of multi-task models to adversarial attacks, which is critical for real-world applications, but it is incremental as it refines prior claims.
The paper investigates how design choices affect adversarial robustness in multi-task deep learning networks, finding that adding auxiliary tasks or weighing them can create a false sense of robustness, and identifies task selection in the loss function as a key factor for improving robustness.
Vulnerability to adversarial attacks is a well-known weakness of Deep Neural networks. While most of the studies focus on single-task neural networks with computer vision datasets, very little research has considered complex multi-task models that are common in real applications. In this paper, we evaluate the design choices that impact the robustness of multi-task deep learning networks. We provide evidence that blindly adding auxiliary tasks, or weighing the tasks provides a false sense of robustness. Thereby, we tone down the claim made by previous research and study the different factors which may affect robustness. In particular, we show that the choice of the task to incorporate in the loss function are important factors that can be leveraged to yield more robust models.