ML Anthology

Towards One Shot Search Space Poisoning in Neural Architecture Search (Student Abstract)

Nayan Saxena, Robert Wu, Rohan Jain
AAAI 2022 pp. 13043-13044
doi:10.1609/AAAI.V36I11.21658 /aaai/2022/saxena2022aaai-one/

Abstract

We evaluate the robustness of a Neural Architecture Search (NAS) algorithm known as Efficient NAS (ENAS) against data agnostic poisoning attacks on the original search space with carefully designed ineffective operations. We empirically demonstrate how our one shot search space poisoning approach exploits design flaws in the ENAS controller to degrade predictive performance on classification tasks. With just two poisoning operations injected into the search space, we inflate prediction error rates for child networks upto 90% on the CIFAR-10 dataset.

PDF AAAI Semantic Scholar

Cite

Text

Saxena et al. "Towards One Shot Search Space Poisoning in Neural Architecture Search (Student Abstract)." AAAI Conference on Artificial Intelligence, 2022. doi:10.1609/AAAI.V36I11.21658

Markdown

[Saxena et al. "Towards One Shot Search Space Poisoning in Neural Architecture Search (Student Abstract)." AAAI Conference on Artificial Intelligence, 2022.](https://mlanthology.org/aaai/2022/saxena2022aaai-one/) doi:10.1609/AAAI.V36I11.21658

BibTeX

@inproceedings{saxena2022aaai-one,
  title     = {{Towards One Shot Search Space Poisoning in Neural Architecture Search (Student Abstract)}},
  author    = {Saxena, Nayan and Wu, Robert and Jain, Rohan},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2022},
  pages     = {13043-13044},
  doi       = {10.1609/AAAI.V36I11.21658},
  url       = {https://mlanthology.org/aaai/2022/saxena2022aaai-one/}
}