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 2310.12665cs.CRcs.LG
收藏

SecurityNet:评估公共模型上的机器学习漏洞

SecurityNet: Assessing Machine Learning Vulnerabilities on Public Models

CISPA Helmholtz Center for Information Security(CISPA亥姆霍兹信息安全中心)

Boyang Zhang, Zheng Li, Ziqing Yang, Xinlei He, Michael Backes, Mario Fritz, Yang Zhang

AI总结:

提出使用包含910个公开图像分类模型的SecurityNet数据库评估机器学习攻击与防御,发现这些方法在公共模型上的表现与自训练模型存在显著差异。

摘要:

尽管先进的机器学习(ML)模型已被部署于众多实际应用中,但先前的研究表明这些模型存在安全与隐私漏洞。该领域已开展了大量实证研究。然而,大多数实验都是在安全研究人员自己训练的目标ML模型上进行的。由于训练具有复杂架构的先进模型需要极高的计算资源,研究人员通常选择在典型实验数据集上,使用相对简单的架构训练少量目标模型。我们认为,为了全面理解ML模型的漏洞,应在为各种目的(不仅是为了评估ML攻击和防御)而训练的大量模型上进行实验。为此,我们提出使用从互联网上获取的带有权重的公开可用模型(公共模型)来评估ML模型的攻击和防御。我们建立了一个名为SecurityNet的数据库,包含910个带标注的图像分类模型。随后,我们分析了几种代表性攻击/防御的有效性,包括模型窃取攻击、成员推理攻击以及在这些公共模型上的后门检测。我们的评估实证表明,与自训练模型相比,这些攻击/防御在公共模型上的性能可能会有显著差异。我们与研究社区共享SecurityNet,并倡导研究人员未来在公共模型上进行实验,以更好地证明其提出方法的有效性。

英文摘要:

While advanced machine learning (ML) models are deployed in numerous real-world applications, previous works demonstrate these models have security and privacy vulnerabilities. Various empirical research has been done in this field. However, most of the experiments are performed on target ML models trained by the security researchers themselves. Due to the high computational resource requirement for training advanced models with complex architectures, researchers generally choose to train a few target models using relatively simple architectures on typical experiment datasets. We argue that to understand ML models' vulnerabilities comprehensively, experiments should be performed on a large set of models trained with various purposes (not just the purpose of evaluating ML attacks and defenses). To this end, we propose using publicly available models with weights from the Internet (public models) for evaluating attacks and defenses on ML models. We establish a database, namely SecurityNet, containing 910 annotated image classification models. We then analyze the effectiveness of several representative attacks/defenses, including model stealing attacks, membership inference attacks, and backdoor detection on these public models. Our evaluation empirically shows the performance of these attacks/defenses can vary significantly on public models compared to self-trained models. We share SecurityNet with the research community. and advocate researchers to perform experiments on public models to better demonstrate their proposed methods' effectiveness in the future.

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。

↑