Wild Patterns Reloaded: A Survey of Machine Learning Security against Training Data Poisoning
ACM Computing Surveys · 31.12.2023 · 55 (13s) · 1-39
DOI: 10.1145/3585385 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Computing Surveys, online erschienen am 13.07.2023. Verfasst von Antonio Emanuele Cinà und 9 weiteren Autor:innen. Die Arbeit wurde bisher 153-mal zitiert (Crossref) und nennt 215 Literaturangaben. Gefördert von: PRIN 2017 project RexLearn, Italian Ministry of Education, University and Research, EU’s Horizon Europe research and innovation program under the project ELSA.
Abstract (Originalsprache)
The success of machine learning is fueled by the increasing availability of computing power and large training datasets. The training data is used to learn new models or update existing ones, assuming that it is sufficiently representative of the data that will be encountered at test time. This assumption is challenged by the threat of poisoning, an attack that manipulates the training data to compromise the model’s performance at test time. Although poisoning has been acknowledged as a relevant threat in industry applications, and a variety of different attacks and defenses have been proposed so far, a complete systematization and critical review of the field is still missing. In this survey, we provide a comprehensive systematization of poisoning attacks and defenses in machine learning, reviewing more than 100 papers published in the field in the past 15 years. We start by categorizing the current threat models and attacks and then organize existing defenses accordingly. While we focus mostly on computer-vision applications, we argue that our systematization also encompasses state-of-the-art attacks and defenses for other data modalities. Finally, we discuss existing resources for research in poisoning and shed light on the current limitations and open research questions in this research field.
Förderung
- PRIN 2017 project RexLearn
- Italian Ministry of Education, University and Research
- EU’s Horizon Europe research and innovation program under the project ELSA
- “TrustML: Towards Machine Learning that Humans Can Trust,”
- Fondazione di Sardegna; the NRRP MUR program funded by the EU - NGEU under the project SERICS
- COMET Programme managed by FFG in the COMET Module S3AI
Zitieren
APA
Cinà, A. E., Grosse, K., Demontis, A., Vascon, S., Zellinger, W., Moser, B. A., Oprea, A., Biggio, B., Pelillo, M., & Roli, F. (2023). Wild Patterns Reloaded: A Survey of Machine Learning Security against Training Data Poisoning. ACM Computing Surveys, 55(13s), 1-39. https://doi.org/10.1145/3585385
Vancouver
Cinà AE, Grosse K, Demontis A, Vascon S, Zellinger W, Moser BA, et al. Wild Patterns Reloaded: A Survey of Machine Learning Security against Training Data Poisoning. ACM Computing Surveys. 2023;55(13s):1-39. doi:10.1145/3585385
BibTeX
@article{Cin2023,
author = {Cinà, Antonio Emanuele and Grosse, Kathrin and Demontis, Ambra and Vascon, Sebastiano and Zellinger, Werner and Moser, Bernhard A. and Oprea, Alina and Biggio, Battista and Pelillo, Marcello and Roli, Fabio},
title = {Wild Patterns Reloaded: A Survey of Machine Learning Security against Training Data Poisoning},
journal = {ACM Computing Surveys},
year = {2023},
volume = {55},
number = {13s},
pages = {1--39},
doi = {10.1145/3585385}
}
EUVIMED