Research map / Adversarial robustness and security / Backdoor and poisoning
Poison-then-Hide: Finetuning-Activated Backdoor Attack on Pretrained Vision Encoders
Qixuan Jin, Abinitha Gourabathina, Vinith Suriyakumar, Walter Gerych, Marzyeh Ghassemi
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a finetuning-activated backdoor attack on pretrained vision encoders that stays dormant until downstream full-model finetuning reactivates it.
Problem
Existing finetuning-activated backdoor attacks assume limited domain shift or frozen encoder layers, so they often fail under full-model finetuning of pretrained vision encoders.
Approach
A poison-then-hide strategy: the backdoor is planted so it is concealed in the released pretrained encoder and becomes activated only when the downstream model is fully finetuned. Details are not available beyond the abstract.
Closest papers by meaning
- Hidden Trigger Backdoor Attacks (AAAI 2020 · similarity 0.91)
- Unveiling and Mitigating Backdoor Vulnerabilities based on Unlearning Weight Changes and Backdoor Activeness (NeurIPS 2024 · similarity 0.90)
- Demystifying Poisoning Backdoor Attacks from a Statistical Perspective (ICLR 2024 · similarity 0.89)
- A Set of Generalized Components to Achieve Effective Poison-only Clean-label Backdoor Attacks with Collaborative Sample Selection and Triggers (NeurIPS 2025 · similarity 0.89)
- Wicked Oddities: Selectively Poisoning for Effective Clean-Label Backdoor Attacks (ICLR 2025 · similarity 0.89)
- Privacy Backdoors: Enhancing Membership Inference through Poisoning Pre-trained Models (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
