Skip to content
Search
paper2024ReviewedOpen access

Poisoning Web-Scale Training Datasets is Practical

Nicholas Carlini, Matthew Jagielski, Christopher A. Choquette-Choo, Daniel Paleka, Will Pearce, Hyrum Anderson, Andreas Terzis, Kurt Thomas, Florian Tramer

IEEE S&P 2024

View resourcePDF

Abstract

Demonstrates practical attacks to poison web-scale datasets like LAION by purchasing expired domains, affecting 0.01% of a dataset for under $60.

Categories

Data PoisoningSupply Chain Attacks
#web-scale#practical-attack#domain-purchase#dataset-poisoning

Framework mappings

OWASP Top 10 for LLM Applications
  • LLM03Supply Chain
  • LLM04Data and Model Poisoning
MITRE ATLAS
  • AML.T0020Poison Training Data

Cite

@inproceedings{carlini2024poisoning,
  title = {{Poisoning Web-Scale Training Datasets is Practical}},
  author = {Nicholas Carlini and Matthew Jagielski and Christopher A. Choquette-Choo and Daniel Paleka and Will Pearce and Hyrum Anderson and Andreas Terzis and Kurt Thomas and Florian Tramer},
  year = {2024},
  booktitle = {IEEE S\&P 2024},
  eprint = {2302.10149},
  archivePrefix = {arXiv},
  doi = {10.1109/sp54263.2024.00179},
  url = {https://arxiv.org/abs/2302.10149}
}