Low-Loss Subspace Compression for Clean Gains Against Multi-Agent Backdoor Attacks

Abstract

Recent exploration of the multi-agent backdoor attack demonstrated the backfiring effect, a natural defense against backdoor attacks where backdoored inputs are randomly classified. This yields a side-effect of low accuracy w.r.t. clean labels, which motivates this paper's work on the construction of multi-agent backdoor defenses that maximize accuracy w.r.t. clean labels and minimize that of poison labels. Founded upon agent dynamics and low-loss subspace construction, we contribute three defenses that yield improved multi-agent backdoor robustness.

Cite

Text

Datta and Shadbolt. "Low-Loss Subspace Compression for Clean Gains Against Multi-Agent Backdoor Attacks." ICML 2022 Workshops: AI4ABM, 2022.

Markdown

[Datta and Shadbolt. "Low-Loss Subspace Compression for Clean Gains Against Multi-Agent Backdoor Attacks." ICML 2022 Workshops: AI4ABM, 2022.](https://mlanthology.org/icmlw/2022/datta2022icmlw-lowloss/)

BibTeX

@inproceedings{datta2022icmlw-lowloss,
  title     = {{Low-Loss Subspace Compression for Clean Gains Against Multi-Agent Backdoor Attacks}},
  author    = {Datta, Siddhartha and Shadbolt, Nigel},
  booktitle = {ICML 2022 Workshops: AI4ABM},
  year      = {2022},
  url       = {https://mlanthology.org/icmlw/2022/datta2022icmlw-lowloss/}
}