Updated 'PGD' columns of Table 1: numbers reported earlier for this column were (100 - accuracy) instead of attack success rates. Observations and conclusions remain unchanged
Despite vast research in adversarial examples, the root causes of model susceptibility are not well understood. Instead of looking at attack-specific robustness, we propose a notion that evaluates the sensitivity of individual neurons in terms of how robust the model's output is to direct perturbations of that neuron's output. Analyzing models from this perspective reveals distinctive characteristics of standard as well as adversarially-trained robust models, and leads to several curious results. In our experiments on CIFAR-10 and ImageNet, we find that attacks using a loss function that targets just a single sensitive neuron find adversarial examples nearly as effectively as ones that target the full model. We analyze the properties of these sensitive neurons to propose a regularization term that can help a model achieve robustness to a variety of different perturbation constraints while maintaining accuracy on natural data distributions. Code for all our experiments is available at https://github.com/iamgroot42/sauron .
DANCin SEQ2SEQ: Fooling Text Classifiers with Adversarial Text Example Generation
Stanford University(斯坦福大学)
机器学习模型功能强大,但也容易出错。生成对抗性示例-故意制作的输入,以导致模型错误分类或其他错误-可以对模型假设和漏洞产生重要的洞察力。尽管最近针对图像分类器的对抗性示例生成的工作很重要,但探索文本分类器的对抗性示例生成的工作相对较少;此外,许多现有的对抗性示例生成算法需要完全访问目标模型参数,这使得它们对于许多现实世界的攻击来说不切实际。在这项工作中,我们引入了DANC in SEQ 2SEQ,这是一种受GAN启发的算法,用于生成对抗性文本示例,主要针对黑盒文本分类器。我们将对抗性文本示例生成重新定义为一个强化学习问题,并证明我们的算法为在真实攻击场景中生成语义上有意义的对抗性文本示例提供了初步但有希望的步骤。
One Attack to Fool Them All: Highly Transferable Black-Box Adversarial Attacks on Frontier MLLMs
State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences(中国科学院计算技术研究所人工智能安全国家重点实验室);University of Chinese Academy of Sciences(中国科学院大学)
Dense Associative Memory is Robust to Adversarial Inputs
Simons Center for Systems Biology, Institute for Advanced Study(高等研究院西蒙斯系统生物学中心);Princeton Neuroscience Institute, Princeton University(普林斯顿大学普林斯顿神经科学研究所)
Universal Adversarial Backdoor Attacks to Fool Vertical Federated Learning in Cloud-Edge Collaboration
Fudan University(复旦大学);Engineering Research Center of Cyber Security Auditing and Monitoring, Ministry of Education(教育部网络安全审计与监控工程研究中心);Shanghai Blockchain Engineering Research Center(上海区块链工程研究中心);Institute of Financial Technology, Fudan University(复旦大学金融科技研究院)
针对云边协同 VFL 二分类场景,提出无需额外数据的 UAB 干净标签后门攻击,通过交替优化通用触发器与子问题参数注入后门,在两个数据集上达到最高 100% 攻击性能且不损害主任务。
Assessing Neural Network Robustness via Adversarial Pivotal Tuning
University of Copenhagen(哥本哈根大学);Pioneer Centre for AI(先锋人工智能中心);KTH Stockholm(瑞典皇家理工学院);Helmholtz AI(亥姆霍兹人工智能);Hebrew University of Jerusalem(耶路撒冷希伯来大学)