Hiding Behind Backdoors: Self-Obfuscation Against Generative Models
On the map: LLM safety and security → Backdoor and data poisoning attacks
Abstract (arXiv)
Attack vectors that compromise machine learning pipelines in the physical world have been demonstrated in recent research, from perturbations to architectural components. Building on this work, we illustrate the self-obfuscation attack: attackers target a pre-processing model in the system, and poison the training set of generative models to obfuscate a specific class during inference. Our contribution is to describe, implement and evaluate a generalized attack, in the hope of raising awareness regarding the challenge of architectural robustness within the machine learning community.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Backdoor and data poisoning attacks, one of 12 directions in LLM safety and security. Builds on 27 of its 41 references that are in the map. First posted on arXiv 24 Jan 2022.
Builds on
27 of 41 references.
References · 27
- 2014Microsoft COCO: Common Objects in Contextcited by 17,498
- 2014Explaining and Harnessing Adversarial Examplescited by 7,287
- 2016Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Networkcited by 1,898
- 2017Enhanced Deep Residual Networks for Single Image Super-Resolutioncited by 1,543
- 2017Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoningcited by 1,012
- 2018Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networkscited by 500
- 2019EnlightenGAN: Deep Light Enhancement without Paired Supervisioncited by 466
- 2018Spectral Signatures in Backdoor Attackscited by 457
- 2018Detecting Backdoor Attacks on Deep Neural Networks by Activation Clusteringcited by 429
- 2017Data Augmentation Generative Adversarial Networkscited by 278
- 2019Fooling automated surveillance cameras: adversarial patches to attack person detectioncited by 268
- 2019Joint Discriminative and Generative Learning for Person Re-identificationcited by 211
15 more
- 2019Making an Invisibility Cloak: Real World Adversarial Attacks on Object Detectorscited by 155
- 2019Transferable Clean-Label Poisoning Attacks on Deep Neural Netscited by 129
- 2020Backdoor Attacks Against Deep Learning Systems in the Physical Worldcited by 129
- 2019TABOR: A Highly Accurate Approach to Inspecting and Restoring Trojan Backdoors in AI Systemscited by 120
- 2021Backdoor Attack in the Physical Worldcited by 82
- 2018Wide Activation for Efficient and Accurate Image Super-Resolutioncited by 64
- 2019ChromaGAN: Adversarial Picture Colorization with Semantic Class Distributioncited by 60
- 2018A Hybrid Model for Identity Obfuscation by Face Replacementcited by 43
- 2017Unsupervised Diverse Colorization via Generative Adversarial Networkscited by 30
- 2020Backdooring and Poisoning Neural Networks with Image-Scaling Attackscited by 28
- 2020BAAAN: Backdoor Attacks Against Autoencoder and GAN-Based Machine Learning Modelscited by 23
- 2019Poison as a Cure: Detecting & Neutralizing Variable-Sized Backdoor Attacks in Deep Neural Networkscited by 21
- 2017A Berkeley View of Systems Challenges for AIcited by 15
- 2019Contamination Attacks and Mitigation in Multi-Party Machine Learningcited by 13
- 2019A Utility-Preserving GAN for Face Obscurationcited by 5
Built on by
Citing papers per year: 2022: 2.
2 papers.