arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2110.04571cs.LGcs.CR
收藏

扩大后门,让更多的攻击者进入

Widen The Backdoor To Let More Attackers In

University of Oxford(牛津大学)

Siddhartha Datta, Giulio Lovisotto, Ivan Martinovic, Nigel Shadbolt

摘要:

随着协作学习和数据汇集外包变得越来越普遍,试图操纵学习过程的恶意行为者(或代理人)在相互竞争时面临额外的障碍。在后门攻击中,对手试图通过将恶意样本引入训练数据来毒害模型,对手必须考虑到额外后门攻击者的存在可能会阻碍他们自己的后门的成功。在本文中,我们研究了多代理后门攻击的场景,其中多个非合谋攻击者在共享数据集中制作并插入触发样本,该数据集由模型(防御者)用于学习任务。我们发现了一个明显的逆火现象:增加攻击者的数量减少了每个攻击者的攻击成功率(ASR)。然后,我们利用这种现象来最大限度地减少攻击者的集体ASR,并通过(i)人工增加攻击者的数量,以及(ii)索引以从模型中删除攻击者的子数据集进行推理,从而最大限度地提高防御者的鲁棒性准确性,从而提出2种防御方法。

英文摘要:

As collaborative learning and the outsourcing of data collection become more common, malicious actors (or agents) which attempt to manipulate the learning process face an additional obstacle as they compete with each other. In backdoor attacks, where an adversary attempts to poison a model by introducing malicious samples into the training data, adversaries have to consider that the presence of additional backdoor attackers may hamper the success of their own backdoor. In this paper, we investigate the scenario of a multi-agent backdoor attack, where multiple non-colluding attackers craft and insert triggered samples in a shared dataset which is used by a model (a defender) to learn a task. We discover a clear backfiring phenomenon: increasing the number of attackers shrinks each attacker's attack success rate (ASR). We then exploit this phenomenon to minimize the collective ASR of attackers and maximize defender's robustness accuracy by (i) artificially augmenting the number of attackers, and (ii) indexing to remove the attacker's sub-dataset from the model for inference, hence proposing 2 defenses.

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。

↑