mlsanityData to 14 Oct 2026

Efficient Black-Box Adversarial Attacks on Neural Text Detectors

arXiv 2311.01873 ↗First posted 3 Nov 2023cs.CLcited by 3

Not on the map. Why?

Abstract (arXiv)

Neural text detectors are models trained to detect whether a given text was generated by a language model or written by a human. In this paper, we investigate three simple and resource-efficient strategies (parameter tweaking, prompt engineering, and character-level mutations) to alter texts generated by GPT-3.5 that are unsuspicious or unnoticeable for humans but cause misclassification by neural text detectors. The results show that especially parameter tweaking and character-level mutations are effective strategies.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 12 of its 20 references that are in the map. First posted on arXiv 3 Nov 2023.

Builds on

12 of 20 references.

References · 12

  1. 2014Explaining and Harnessing Adversarial Examplescited by 7,364
  2. 2013Intriguing properties of neural networkscited by 5,155
  3. 2019Is BERT Really Robust? A Strong Baseline for Natural Language Attack on Text Classification and Entailmentcited by 671
  4. 2023Can AI-Generated Text be Reliably Detected?cited by 386
  5. 2018Black-box Generation of Adversarial Text Sequences to Evade Deep Learning Classifierscited by 345
  6. 2023GPT detectors are biased against non-native English writerscited by 178
  7. 2020Automatic Detection of Machine Generated Text: A Critical Surveycited by 166
  8. 2022Detection of Word Adversarial Examples in Text Classification: Benchmark and Baseline via Robust Density Estimationcited by 32
  9. 2022Adversarial Robustness of Neural-Statistical Features in Detection of Generative Transformerscited by 25
  10. 2021Untargeted, Targeted and Universal Adversarial Attacks and Defenses on Time Seriescited by 21
  11. 2023Mutation-Based Adversarial Attacks on Neural Text Detectorscited by 9
  12. 2023TextDefense: Adversarial Text Detection based on Word Importance Entropycited by 9

Built on by

Cited by year

Citing papers per year: 2024: 2; 2025: 1.

3 papers.

  1. 2024RAID: A Shared Benchmark for Robust Evaluation of Machine-Generated Text Detectorscited by 163
  2. 2024LLM Detectors Still Fall Short of Real World: Case of LLM-Generated Short News-Like Postscited by 4
  3. 2025How Sampling Affects the Detectability of Machine-written texts: A Comprehensive Studycited by 4