Hidden Poison: Machine Unlearning Enables Camouflaged Poisoning Attacks

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Di, Jimmy Z., Douglas, Jack, Acharya, Jayadev, Kamath, Gautam, Sekhari, Ayush
Format: Preprint
Veröffentlicht: 2022
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866916341737324544
author Di, Jimmy Z.
Douglas, Jack
Acharya, Jayadev
Kamath, Gautam
Sekhari, Ayush
author_facet Di, Jimmy Z.
Douglas, Jack
Acharya, Jayadev
Kamath, Gautam
Sekhari, Ayush
contents We introduce camouflaged data poisoning attacks, a new attack vector that arises in the context of machine unlearning and other settings when model retraining may be induced. An adversary first adds a few carefully crafted points to the training dataset such that the impact on the model's predictions is minimal. The adversary subsequently triggers a request to remove a subset of the introduced points at which point the attack is unleashed and the model's predictions are negatively affected. In particular, we consider clean-label targeted attacks (in which the goal is to cause the model to misclassify a specific test point) on datasets including CIFAR-10, Imagenette, and Imagewoof. This attack is realized by constructing camouflage datapoints that mask the effect of a poisoned dataset.
format Preprint
id arxiv_https___arxiv_org_abs_2212_10717
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Hidden Poison: Machine Unlearning Enables Camouflaged Poisoning Attacks
Di, Jimmy Z.
Douglas, Jack
Acharya, Jayadev
Kamath, Gautam
Sekhari, Ayush
Machine Learning
Artificial Intelligence
Cryptography and Security
Computers and Society
We introduce camouflaged data poisoning attacks, a new attack vector that arises in the context of machine unlearning and other settings when model retraining may be induced. An adversary first adds a few carefully crafted points to the training dataset such that the impact on the model's predictions is minimal. The adversary subsequently triggers a request to remove a subset of the introduced points at which point the attack is unleashed and the model's predictions are negatively affected. In particular, we consider clean-label targeted attacks (in which the goal is to cause the model to misclassify a specific test point) on datasets including CIFAR-10, Imagenette, and Imagewoof. This attack is realized by constructing camouflage datapoints that mask the effect of a poisoned dataset.
title Hidden Poison: Machine Unlearning Enables Camouflaged Poisoning Attacks
topic Machine Learning
Artificial Intelligence
Cryptography and Security
Computers and Society
url https://arxiv.org/abs/2212.10717