Pith. sign in

REVIEW 9 cited by

Concealed Data Poisoning Attacks on NLP Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2010.12563 v2 pith:6434FX4X submitted 2020-10-23 cs.CL

classification cs.CL
keywords modelattackdatapoisonpredictionsattackscoffeeconcealed
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

Adversarial attacks alter NLP model predictions by perturbing test-time inputs. However, it is much less understood whether, and how, predictions can be manipulated with small, concealed changes to the training data. In this work, we develop a new data poisoning attack that allows an adversary to control model predictions whenever a desired trigger phrase is present in the input. For instance, we insert 50 poison examples into a sentiment model's training set that causes the model to frequently predict Positive whenever the input contains "James Bond". Crucially, we craft these poison examples using a gradient-based procedure so that they do not mention the trigger phrase. We also apply our poison attack to language modeling ("Apple iPhone" triggers negative generations) and machine translation ("iced coffee" mistranslated as "hot coffee"). We conclude by proposing three defenses that can mitigate our attack at some cost in prediction accuracy or extra human annotation.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Pretraining Data Can Be Poisoned through Computational Propaganda

    cs.AI 2026-07 conditional novelty 7.0 of 10

    Public comment sections are a measurable, low-cost vector for injecting poison into web-scale language-model pretraining data.

  2. Task-Agnostic Language Model Watermarking via High Entropy Passthrough Layers

    cs.CL 2024-12 reject novelty 6.0 of 10

    A backdoor watermark for LLMs using passthrough layers trained to output high-entropy text on a private key, with near-perfect extraction in benign settings but with layer-removal robustness contradicted by the paper'...

  3. When Backdoors Speak: Understanding LLM Backdoor Attacks Through Model-Generated Explanations

    cs.CR 2024-11 conditional novelty 6.0 of 10

    Backdoored LLMs produce more diverse, less coherent explanations on triggered inputs, and this difference can be used to detect the backdoor.

  4. A Red Teaming Framework for Large Language Models: A Case Study on Faithfulness Evaluation

    cs.CL 2026-06 unverdicted novelty 5.0 of 10

    Introduces a multi-role red teaming framework using attacker and jury models that increases attack success rates by up to 7.9% on LLM faithfulness in question-answering tasks.

  5. A Systematic Review of Poisoning Attacks Against Large Language Models

    cs.CR 2025-06 conditional novelty 5.0 of 10

    A systematic review that organizes 65 LLM poisoning papers into a threat model with four attack specifications and generalized metrics.

  6. A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deployment

    cs.CR 2025-04 conditional novelty 5.0 of 10

    A large collaborative survey organizes LLM and LLM-agent safety issues into a full-stack lifecycle framework from data preparation to deployment.

  7. Towards Data Governance of Frontier AI Models

    cs.AI 2024-12 conditional novelty 5.0 of 10

    Training data can serve as a governance lever for frontier AI through five proposed mechanisms: canary tokens, mandatory filtering, dataset reporting, data security, and know-your-customer rules.

  8. LLM Security: Vulnerabilities, Attacks, Defenses, and Countermeasures

    cs.CR 2025-05 conditional novelty 3.0 of 10

    This survey categorizes attacks on large language models by lifecycle phase and maps them to prevention and detection defenses, concluding that only a few defenses are highly effective.

  9. A Survey on Data Security in Large Language Models

    cs.CR 2025-08 conditional novelty 2.0 of 10

    A survey of data security risks in LLMs that organizes threats, defenses, and evaluation datasets, with notable factual errors in its tables.

Pith tools