FlipGuard: Defending Large Language Models Against Quantization-Conditioned Backdoor Attacks

Abstract

Model quantization is essential for the efficient deployment of Large Language Models (LLMs), but introduces a critical vulnerability: Quantization-Conditioned Backdoor (QCB) attacks. In these attacks, malicious behaviors remain dormant in full-precision models and activate only after specific quantization distortions, bypassing standard security audits. To mitigate this, we introduce FlipGuard, a proactive defense framework that selectively perturbs model weights prior to quantization. By break

Categories

Framework mappings

Suggested from the entry's categories.

@misc{zheng2026flipguard,
  title = {{FlipGuard: Defending Large Language Models Against Quantization-Conditioned Backdoor Attacks}},
  author = {Aoying Zheng and Anqi Du and Zizhuang Deng and Yuxuan Chen},
  year = {2026},
  month = jun,
  eprint = {2606.28962},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2606.28962}
}