With Great Training Comes Great Vulnerability: Practical Attacks against Transfer Learning

By Bolun Wang, Yuanshun Yao, Bimal Viswanath, Haitao Zheng, Ben Y. Zhao. 2018

126 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks (2019, background): “These can be applied to DNNs during inference [38], [39], [40], [41], [42].”
A Survey on Negative Transfer (2020, background): “For example, the white-box teacher model, black-box student model and TL parameters can be affected by evasion attacks [121]–[123], and source data by backdoor attacks [124]–[126].”
Understanding and Controlling User Linkability in Decentralized Learning (2018, background): “ tasks such as autonomous driving, fraud detection, and medical diagnosis. Attacks against such ML models can be targeted towards compromising the integrity of the model (such as by evasion 2 attacks [5,12,16,29,57,70,73,77,81]), or its privacy and conﬁdentiality. Our focus is on the latter, since M”
Poisoning the Unlabeled Dataset of Semi-Supervised Learning (2021, background): “For example, if it is public that a model has been fine-tuned from a particular ImageNet model, it becomes easier to generate adversarial examples on the fine-tuned model [63].”
Gradient-Leaks: Understanding and Controlling Deanonymization in Federated Learning (2018, background): “Attacks against such ML models can be targeted towards compromising the integrity of the model (such as by evasion attacks [5,13,18,32,63,76,79,84,87]), or its privacy and con-ﬁdentiality .”
Better the Devil you Know: An Analysis of Evasion Attacks using Out-of-Distribution Adversarial Examples (2019, background): “ttacks We focus on test phase or evasion attacks. 2.2.1 Adversarialexamplesgeneratedfromin-distributiondata. Evasion attacks have been demonstrated to be highly successful for a number of classifiers [13, 18, 40, 45, 73, 88, 97, 110, 113, 114, 118]. All of these aim to modify benign, in-distribution”
Updates-Leak: Data Set Inference and Reconstruction Attacks in Online Learning (2019, unknown): “Besides the above, there exist a wide range of other attacks and defenses on machine learning models [4,5,8,9,13,14,16, 17, 22, 23, 25, 26, 35, 41, 43, 44, 46, 48–50, 52, 54–56]”
Fawkes: Protecting Privacy against Unauthorized Deep Learning Models (2020, methodology): “lculated using the DSSIM (Structural Dis-Similarity Index) [61, 62]. Different from the L p distance used in previous work [9,25,43], DSSIMhasgainedpopularityasameasureofuser-perceived imagedistortion[23,28,59].Boundingcloakgenerationwith this metric ensures that cloaked versions of images are visua”
Latent Backdoor Attacks on Deep Neural Networks (2019, unknown): “These can be applied to models during inference [9, 25, 29, 37, 50].”