arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2606.09548cs.CRcs.AI
收藏

基于比特翻转链的联邦模型自适应中毒攻击

Model Poisoning Against Federated Model Adaptation with Chain of Bit-Flips

CEA-Leti, Mines Saint-Etienne, Equipe Commune SAS(CEA-莱蒂, Mines圣艾蒂安, 共同团队SAS);Univ. Grenoble Alpes, CEA-Leti(格勒诺布尔阿尔卑斯大学, CEA-莱蒂);Mines Saint-Etienne, CEA-Leti, Centre CMP, Equipe commune SAS(Mines圣艾蒂安, CEA-莱蒂, CMP中心, 共同团队SAS)

Bastien Vuillod, Kevin Hector, Pierre-Alain Moellic, Jean-Max Dutertre, Olivier Potin

AI总结:

提出一种结合硬件故障攻击的模型中毒方法,在联邦学习训练阶段通过比特翻转注入后门,实现任务无关的后门攻击,在ResNet-18上仅需少量故障即可达到94%攻击成功率。

摘要:

联邦学习允许一组客户端在不共享本地训练数据的情况下共同训练全局模型。将训练责任交给去中心化的参与者可能导致中毒攻击:由恶意第三方控制的客户端可能毒化训练数据集,在神经网络中安装后门。在联邦学习中,这些后门攻击仅依赖算法方法,然而,硬件故障威胁(如Rowhammer)的最新进展拓宽了整体攻击面。在联邦模型自适应的背景下,我们引入了一种针对联邦学习系统的新型后门攻击类别,该攻击基于硬件故障攻击的模型中毒。更准确地说,我们提出了一种任务无关的后门攻击,通过在联邦训练期间诱导单个本地模型参数中的硬件故障(比特翻转)来植入后门。后门是在之前的离线阶段从联邦系统最初使用的预训练模型中精心制作的。我们的结果表明,后门可以成功应用于不同类型的模型和数据集。通常,每个恶意客户端出现最多10次故障,且总共出现19次故障,就足以在ResNet-18上达到94%的攻击成功率。最后,我们讨论了攻击潜在防御的实用性和鲁棒性,同时考虑了Rowhammer的实际约束,这是此类威胁的首选攻击向量。

英文摘要:

Federated Learning (FL) allows a set of clients to collectively train a global model without sharing local training data. Giving the responsibility of the training to decentralized actors may lead to poisoning attacks: clients controlled by malicious third party potentially poison the training dataset to install a backdoor in neural networks. In FL, these backdoor attacks rely solely on algorithmic approach, however, recent advances in hardware faults threats (e.g, Rowhammer) have widen the overall attack surface. In the context of federated model adaptation, we introduce a novel category of backdoor attack against FL systems that relies on model poisoning based on hardware-fault attacks. More precisely, we propose a task-agnostic backdoor attack that is implanted during the FL training time by inducing hardware faults (bit-flips) in parameters of a single local model. The backdoor is crafted during a previous offline phase from the pretrained model initially used by the FL system. Our results show that a backdoor can be successfully applied on different type of models and datasets. Typically, with up to 10 faults per malicious client occurrence and 19 total occurrences on a ResNet-18 are enough to reach 94% of attack success rate. Finally, we discuss the practicality and the robustness of the attack potential defenses, while putting into perspective the practical constraints of Rowhammer, which is the preferred attack vector for this type of threats.

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。

↑