arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2102.11010cs.LGstat.ML
收藏

对抗攻击下贝叶斯分层推理的弹性

Resilience of Bayesian Layer-Wise Explanations under Adversarial Attacks

University of Trieste(的里雅斯特大学);Saarland University(萨尔大学);University of Edinburgh(爱丁堡大学);SISSA(国际高等研究学院)

Ginevra Carbone, Guido Sanguinetti, Luca Bortolussi

摘要:

我们考虑了在分类任务中基于显著性的解释在对抗攻击下的神经网络预测的稳定性问题。确定性神经网络的显着性解释是非常脆弱的,即使在攻击失败时,即对于不改变分类标签的攻击。我们的经验表明,贝叶斯神经网络提供的解释是相当稳定的对抗扰动下的输入,甚至在直接攻击的解释。通过利用最近的结果,我们还提供了一个理论解释的数据流形的几何形状的这一结果。此外,我们还讨论了网络内部各层输入的高级表示的解释的稳定性。我们的研究结果表明,贝叶斯方法,除了更强大的对抗攻击,有潜力提供更稳定和可解释的评估神经网络预测。

英文摘要:

We consider the problem of the stability of saliency-based explanations of Neural Network predictions under adversarial attacks in a classification task. Saliency interpretations of deterministic Neural Networks are remarkably brittle even when the attacks fail, i.e. for attacks that do not change the classification label. We empirically show that interpretations provided by Bayesian Neural Networks are considerably more stable under adversarial perturbations of the inputs and even under direct attacks to the explanations. By leveraging recent results, we also provide a theoretical explanation of this result in terms of the geometry of the data manifold. Additionally, we discuss the stability of the interpretations of high level representations of the inputs in the internal layers of a Network. Our results demonstrate that Bayesian methods, in addition to being more robust to adversarial attacks, have the potential to provide more stable and interpretable assessments of Neural Network predictions.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑