mlsanityData to 14 Oct 2026

A Linear Approach to Data Poisoning

arXiv 2505.15175 ↗First posted 21 May 2025stat.ML cs.CR cs.LGcited by 2

On the map: LLM safety and security → Backdoor and data poisoning attacks

Abstract (arXiv)

Backdoor and data-poisoning attacks can flip predictions with tiny training corruptions, yet a sharp theory linking poisoning strength, overparameterization, and regularization is lacking. We analyze ridge least squares with an unpenalized intercept in the high-dimensional regime (p,n→∞), (p/n→ c). Targeted poisoning is modelled by shifting a (θ)-fraction of one class by a direction (v) and relabelling. Using resolvent techniques and deterministic equivalents from random matrix theory, we derive closed-form limits for the poisoned score explicit in the model parameters. The formulas yield scaling laws, recover the interpolation threshold as (c→1) in the ridgeless limit, and show that the weights align with the poisoning direction. Synthetic experiments match theory across sweeps of the parameters and MNIST backdoor tests show qualitatively consistent trends. The results provide a tractable framework for quantifying poisoning in linear models.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Backdoor and data poisoning attacks, one of 12 directions in LLM safety and security. Builds on 42 of its 70 references that are in the map. First posted on arXiv 21 May 2025.

Builds on

42 of 70 references.

Closest ancestors · 13

Same direction, newest first

  1. 2025Multi-Faceted Studies on Data Poisoning can Advance LLM Development20 Feb
  2. 2024When Backdoors Speak: Understanding LLM Backdoor Attacks Through Model-Generated Explanations19 Nov
  3. 2024BadEdit: Backdooring large language models by model editing20 Mar
  4. 2024Learning to Poison Large Language Models for Downstream Manipulation21 Feb
  5. 2024Instruction Backdoor Attacks Against Customized LLMs14 Feb
  6. 2024Shadowcast: Stealthy Data Poisoning Attacks Against Vision-Language Models5 Feb
  7. 2024Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training10 Jan
  8. 2023From Trojan Horses to Castle Walls: Unveiling Bilateral Data Poisoning Effects in Diffusion Models4 Nov
  9. 2023PoisonPrompt: Backdoor Attack on Prompt-based Large Language Models19 Oct
  10. 2023BadGPT: Exploring Security Vulnerabilities of ChatGPT via Backdoor Attacks to InstructGPT21 Feb
  11. 2022How to Backdoor Diffusion Models?11 Dec
  12. 2022BadPrompt: Backdoor Attacks on Continuous Prompts27 Nov
1 more
  1. 2022Indiscriminate Poisoning Attacks on Unsupervised Contrastive Learning22 Feb

Other references · 29

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
  4. 2013Intriguing properties of neural networkscited by 5,155
  5. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 3,104
  6. 2018Datasheets for Datasetscited by 1,386
  7. 2017BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chaincited by 1,106
  8. 2018Model Cards for Model Reportingcited by 1,080
  9. 2019Green AIcited by 627
  10. 2018Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networkscited by 519
  11. 2019Surprises in High-Dimensional Ridgeless Least Squares Interpolationcited by 513
  12. 2018Spectral Signatures in Backdoor Attackscited by 484
17 more
  1. 2017Certified Defenses for Data Poisoning Attackscited by 303
  2. 2019Label-Consistent Backdoor Attackscited by 296
  3. 2018Manipulating Machine Learning: Poisoning Attacks and Countermeasures for Regression Learningcited by 225
  4. 2018The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigationcited by 165
  5. 2021Poisoning and Backdooring Contrastive Learningcited by 151
  6. 2015High-Dimensional Asymptotics of Prediction: Ridge Regression and Classificationcited by 148
  7. 2021SPECTRE: Defending Against Backdoor Attacks Using Robust Statisticscited by 130
  8. 2021CRFL: Certifiably Robust Federated Learning against Backdoor Attackscited by 91
  9. 2020On Certifying Robustness against Backdoor Attacks via Randomized Smoothingcited by 77
  10. 2021Handcrafted Backdoors in Deep Neural Networkscited by 63
  11. 2024The Stronger the Diffusion Model, the Easier the Backdoor: Data Poisoning to Induce Copyright Breaches Without Adjusting Finetuning Pipelinecited by 33
  12. 2021Excess Capacity and Backdoor Poisoningcited by 24
  13. 2020Poisoned classifiers are not only backdoored, they are fundamentally brokencited by 15
  14. 2023CBD: A Certified Backdoor Detector Based on Local Dominant Probabilitycited by 13
  15. 2021Scalable logistic regression with crossed random effectscited by 4
  16. 2023Consistent and Scalable Composite Likelihood Estimation of Probit Models with Crossed Random Effectscited by 4
  17. 2024ASPIRe: An Informative Trajectory Planner with Mutual Information Approximation for Target Search and Trackingcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Safety-Efficacy Trade Off: Robustness against Data-Poisoningcited by 1
  2. 2026When Stronger Triggers Backfire: A High-Dimensional Theory of Backdoor Attackscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023What Distributions are Robust to Indiscriminate Poisoning Attacks for Linear Learners?
  2. 2023Exploring the Limits of Model-Targeted Indiscriminate Data Poisoning Attacks
  3. 2025Provable one-poison backdoor attacks on linear models and ReLU neural networks
  4. 2023Activation Gradient based Poisoned Sample Detection Against Backdoor Attacks
  5. 2025On Robustness of Linear Classifiers to Targeted Data Poisoning
  6. 2022Analysis and Detectability of Offline Data Poisoning Attacks on Linear Dynamical Systems
  7. 2026Empirical Evaluation of Data Poisoning Attacks in Supervised Learning
  8. 2022Wild Patterns Reloaded: A Survey of Machine Learning Security against Training Data Poisoning