利用数据排序攻击操纵SGD
Manipulating SGD with Data Ordering Attacks
摘要:
机器学习易受各种攻击。现在已经很好地理解了,通过改变底层数据分布,对手可以毒害用它训练的模型或引入后门。在本文中,我们提出了一类新的训练时间攻击,不需要改变底层数据集或模型架构,而只是改变数据提供给模型的顺序。特别是,我们发现攻击者可以阻止模型学习,或者毒害它来学习攻击者指定的行为。此外,我们发现即使是一个逆序的时期也足以减缓模型学习,甚至重置所有的学习进度。事实上,这里提出的攻击并不特定于模型或数据集,而是针对现代学习过程的随机性。我们广泛评估了我们对计算机视觉和自然语言基准的攻击,发现对手可以破坏模型训练,甚至引入后门。
英文摘要:
Machine learning is vulnerable to a wide variety of attacks. It is now well understood that by changing the underlying data distribution, an adversary can poison the model trained with it or introduce backdoors. In this paper we present a novel class of training-time attacks that require no changes to the underlying dataset or model architecture, but instead only change the order in which data are supplied to the model. In particular, we find that the attacker can either prevent the model from learning, or poison it to learn behaviours specified by the attacker. Furthermore, we find that even a single adversarially-ordered epoch can be enough to slow down model learning, or even to reset all of the learning progress. Indeed, the attacks presented here are not specific to the model or dataset, but rather target the stochastic nature of modern learning procedures. We extensively evaluate our attacks on computer vision and natural language benchmarks to find that the adversary can disrupt model training and even introduce backdoors.
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。