Provably effective detection of effective data poisoning attacks

Authors: Jonathan Gallagher, Yasaman Esfandiari, Callen MacPhee, Michael Warren
Published: 2025-01-21

Source: https://arxiv.org/abs/2501.11795

PDF: https://arxiv.org/pdf/2501.11795

Labels Predicted by AI

ポイズニング バックドア攻撃 実験的検証

Please note that these labels were automatically added by AI. Therefore, they may not be entirely accurate.
For more details, please see the About the Literature Database page.

Abstract

This paper establishes a mathematically precise definition of dataset poisoning attack and proves that the very act of effectively poisoning a dataset ensures that the attack can be effectively detected. On top of a mathematical guarantee that dataset poisoning is identifiable by a new statistical test that we call the Conformal Separability Test, we provide experimental evidence that we can adequately detect poisoning attempts in the real world.