Scene-conditioned spatial-misdirection and duration-inflation backdoors succeed at 2.5–10% poison ratios on multimodal scanpath predictors and resist five adapted defenses.
Fine-tuning is all you need to miti- gate backdoor attacks
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
RogueMerge is a unified attack method that jointly optimizes task vectors to succeed after merging, using stochastic min-max simulation for unknown merging settings and a Taylor-approximated DRO for prompt generalization on generative LLMs.
GaussLock embeds traps targeting position, scale, rotation, opacity, and color in 3D Gaussian models to degrade unauthorized fine-tunes while preserving authorized performance.
Hammer and Anvil framework categorizes backdoors by update deviation δ and shows that principled combinations of Type-1 outlier/robust and Type-2 removal defenses resist full-information adaptive adversaries.
QVec removes quantization-conditioned backdoors by subtracting a structured malicious direction from model weights estimated via one quantization pass and task arithmetic, without retraining or trigger samples.
citing papers explorer
-
Follow My Eyes: Backdoor Attacks on Goal-Directed Scanpath Prediction
Scene-conditioned spatial-misdirection and duration-inflation backdoors succeed at 2.5–10% poison ratios on multimodal scanpath predictors and resist five adapted defenses.
-
RogueMerge: Robust and Unified Attacks against LLM Model Merging
RogueMerge is a unified attack method that jointly optimizes task vectors to succeed after merging, using stochastic min-max simulation for unknown merging settings and a Taylor-approximated DRO for prompt generalization on generative LLMs.
-
Immunizing 3D Gaussian Generative Models Against Unauthorized Fine-Tuning via Attribute-Space Traps
GaussLock embeds traps targeting position, scale, rotation, opacity, and color in 3D Gaussian models to degrade unauthorized fine-tunes while preserving authorized performance.
-
Hammer and Anvil: Toward a Theory of Backdoors in Federated Learning
Hammer and Anvil framework categorizes backdoors by update deviation δ and shows that principled combinations of Type-1 outlier/robust and Type-2 removal defenses resist full-information adaptive adversaries.
-
Quantization as a Malicious Task: Removing Quantization-Conditioned Backdoors via Task Arithmetic
QVec removes quantization-conditioned backdoors by subtracting a structured malicious direction from model weights estimated via one quantization pass and task arithmetic, without retraining or trigger samples.