The DOI matches its Crossref record; 103 of the 111 references it cites were found in a registry or an index.
Poisoning Web-Scale Training Datasets is Practical
Open accessAbstract
Deep learning models are often trained on distributed, web-scale datasets crawled from the internet. In this paper, we introduce two new dataset poisoning attacks that intentionally introduce malicious examples to a model’s performance. Our attacks are immediately practical and could, today, poison 10 popular datasets. Our first attack, split-view poisoning, exploits the mutable nature of internet content to ensure a dataset annotator’s initial view of the dataset differs from the view downloaded by subsequent clients. By exploiting specific invalid trust assumptions, we show how we could have poisoned 0.01% of the LAION-400M or COYO-700M datasets for just $60 USD. Our second attack, frontrunning poisoning, targets web-scale datasets that periodically snapshot crowd-sourced content—such as Wikipedia—where an attacker only needs a time-limited window to inject malicious examples. In light of both attacks, we notify the maintainers of each affected dataset and recommended several low-overhead defenses.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“NIST SP 800-204D [13] provides guidance for integrating software supply chain security controls into DevSecOps CI/CD pipelines, but it does not define AI-specific provenance structures or model assurance mechanisms.”
From AI Supply Chain Security: MBOM-PQC Provenance, PQC Attestation, and a Maturity Model for Quantum-Resistant Assurance (2026) · cites this paper as [13]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
Learning Multiple Layers of Features from Tiny Images
Krizhevsky, A. (2024) · doi:10.57702/zp44cu3g · Cited by 23,889
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Learning Transferable Visual Models From Natural Language Supervision
Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al. (2021) · arXiv · doi:10.48550/arxiv.2103.00020 · Cited by 15,293
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., Zhou, Y., Li, W., & Liu, P. J. (2019) · arXiv · doi:10.48550/arxiv.1910.10683 · Cited by 10,929
Scaling Laws for Neural Language Models
Kaplan, J., McCandlish, S., Henighan, T., Brown, T. B., Chess, B., Child, R., Gray, S., Radford, A., Wu, J., & Amodei, D. (2020) · arXiv · doi:10.48550/arxiv.2001.08361 · Cited by 9,550 · cited 1×
“Deep learning is most effective when applied to large datasets [10, 30].”
BackgroundBuilding a Large Annotated Corpus of English: The Penn Treebank
Marcus, M. (1993) · doi:10.21236/ada273556 · Cited by 7,405
Parkhi, O. M., Vedaldi, A., & Zisserman, A. (2015) · Procedings of the British Machine Vision Conference 2015 · doi:10.5244/c.29.41 · Cited by 5,479
Pointer Sentinel Mixture ModelsPreprint
Merity, S., Xiong, C., Bradbury, J., & Socher, R. (2016) · International Conference on Learning Representations · doi:10.48550/arxiv.1609.07843 · Cited by 5,058
VGGFace2: A Dataset for Recognising Faces across Pose and Age
Cao, Q., Shen, L., Xie, W., Parkhi, O. M., & Zisserman, A. (2018) · 2018 13th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2018) · doi:10.1109/fg.2018.00020 · Cited by 3,032
Research works printed but not linked: 3
- 10.18653/v1/n19–1423 · DOI · Paper page
- Guo, Wiki-40B: Multilingual language model dataset, 2440
- Radford, 2019, Language models are unsupervised multitask learners, OpenAI blog, 1, 9
Cited by
Sorted by citations
Medical large language models are vulnerable to data-poisoning attacks
Alber, D. A., Yang, Z., Alyakin, A., Yang, E., Rai, S., Valliani, A. A., Zhang, J., Rosenbaum, G. R., Amend-Thomas, A. K., Kurland, D. B., et al. (2025) · Nature Medicine · doi:10.1038/s41591-024-03445-1 · Cited by 242
Adversarial machine learning :
Vassilev, A., Oprea, A., Fordyce, A. J., & Anderson, H. S. (2024) · doi:10.6028/nist.ai.100-2e2023 · Cited by 230
Self-citationRethinking machine unlearning for large language models
Liu, S., Yao, Y., Jia, J., Casper, S., Baracaldo, N., Hase, P., Yao, Y., Liu, C. Y., Xu, X., Li, H., et al. (2025) · Nature Machine Intelligence · doi:10.1038/s42256-025-00985-0 · Cited by 374
Nightshade: Prompt-Specific Poisoning Attacks on Text-to-Image Generative Models
Shan, S., Ding, W., Passananti, J., Wu, S. L., Zheng, H.-T., & Zhao, B. Y. (2024) · IEEE Symposium on Security and Privacy (S&P) · doi:10.1109/sp54263.2024.00207 · Cited by 166
Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback
Casper, S., Davies, X., Shi, C., Gilbert, T. K., Scheurer, J., Rando, J., Freedman, R., Korbak, T., Lindner, D., Freire, P., et al. (2023) · arXiv · doi:10.48550/arxiv.2307.15217 · Cited by 101
“However, a similar attack is successful for instruction tuning with very few examples [Wan et al., 2023, Xu et al., 2023a], and poisoning web-scale datasets is possible under realistic assumptions [Carlini et al., 2023a].”
Background
Metadata
- Authors
- Carlini, N., Jagielski, M., Choquette-Choo, C. A., Paleka, D., Pearce, W., Anderson, H., Terzis, A., Thomas, K., & Tramèr, F.
- Affiliations
- Google DeepMind (United Kingdom) (GB)
- ETH Zurich (CH)
- Nvidia (United Kingdom) (GB)
- Robust Intelligence
- Google (United States) (US)
- Venue
- 2024 IEEE Symposium on Security and Privacy (SP)
- Publisher
- IEEE
- Published
- 2024-05-19
- Pages
- 407-425
- ISSN
- 1081-6011 (ISSN-L)
- OpenAlex
- W4402264526
- Cited by
- 138
- References
- 111
- Citations by year
- 2026: 53 · 2025: 44 · 2024: 29
- FWCI
- 37.18
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE Symposium on Security and Privacy (S&P) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Privacy-Preserving Technologies in Data
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Keywords
- deep learning security, Wikipedia, crowdsourced data
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
