ISACL: Internal State Analyzer for Copyrighted Training Data Leakage
Not on the map. Why?
Abstract (arXiv)
Large Language Models (LLMs) have revolutionized Natural Language Processing (NLP) but pose risks of inadvertently exposing copyrighted or proprietary data, especially when such data is used for training but not intended for distribution. Traditional methods address these leaks only after content is generated, which can lead to the exposure of sensitive information. This study introduces a proactive approach: examining LLMs' internal states before text generation to detect potential leaks. By using a curated dataset of copyrighted materials, we trained a neural network classifier to identify risks, allowing for early intervention by stopping the generation process or altering outputs to prevent disclosure. Integrated with a Retrieval-Augmented Generation (RAG) system, this framework ensures adherence to copyright and licensing requirements while enhancing data privacy and ethical standards. Our results show that analyzing internal states effectively mitigates the risk of copyrighted data leakage, offering a scalable solution that fits smoothly into AI workflows, ensuring compliance with copyright regulations while maintaining high-quality text generation. The implementation is available on GitHub.footnotehttps://github.com/changhu73/Internalstatesleakage
Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 38 of its 67 references that are in the map. First posted on arXiv 25 Aug 2025.
Builds on
38 of 67 references.
References · 38
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
- 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
- 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
- 2019RoBERTa: A Robustly Optimized BERT Pretraining Approachcited by 13,372
- 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
- 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,411
- 2020Extracting Training Data from Large Language Modelscited by 2,189
- 2019What Does BERT Look At? An Analysis of BERT's Attentioncited by 1,003
- 2019Defending Against Neural Fake Newscited by 559
- 2021Large Language Models Can Be Strong Differentially Private Learnerscited by 412
- 2023Analyzing Leakage of Personally Identifiable Information in Language Modelscited by 285
- 2022Are Large Pre-Trained Language Models Leaking Your Personal Information?cited by 200
26 more
- 2023Don't Make Your LLM an Evaluation Benchmark Cheatercited by 194
- 2023Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4cited by 125
- 2024LLM Dataset Inference: Did you train on my dataset?cited by 110
- 2021Pretrained Language Models for Text Generation: A Surveycited by 95
- 2021Protecting Intellectual Property of Language Generation APIs with Lexical Watermarkcited by 84
- 2024LLM Internal States Reveal Hallucination Risk Faced With a Querycited by 82
- 2023Offsite-Tuning: Transfer Learning without Full Modelcited by 78
- 2023Are You Copying My Model? Protecting the Copyright of Large Language Models for EaaS via Backdoor Watermarkcited by 67
- 2023DP-Forward: Fine-tuning and Inference on Language Models with Differential Privacy in Forward Passcited by 67
- 2023Cognitive Dissonance: Why Do Language Model Outputs Disagree with Internal Representations of Truthfulness?cited by 62
- 2023"It's a Fair Game", or Is It? Examining How Users Navigate Disclosure Risks and Benefits When Using LLM-Based Conversational Agentscited by 60
- 2022Privately Fine-Tuning Large Language Models with Differential Privacycited by 39
- 2025LessLeak-Bench: A First Investigation of Data Leakage in LLMs Across 83 Software Engineering Benchmarkscited by 37
- 2024CopyBench: Measuring Literal and Non-Literal Reproduction of Copyright-Protected Text in Language Model Generationcited by 35
- 2020Intellectual Property Protection for Deep Learning Models: Taxonomy, Methods, Attacks, and Evaluationscited by 31
- 2023Quantifying Association Capabilities of Large Language Models and Its Implications on Privacy Leakagecited by 28
- 2025Towards Fully Exploiting LLM Internal States to Enhance Knowledge Boundary Perceptioncited by 27
- 2024Noisy Neighbors: Efficient membership inference attacks against LLMscited by 21
- 2023What can we learn from Data Leakage and Unlearning for Law?cited by 15
- 2024Do LLMs Know to Respect Copyright Notice?cited by 13
- 2024Pandora's White-Box: Precise Training Data Detection and Extraction in Large Language Modelscited by 11
- 2024Rethinking Artistic Copyright Infringements in the Era of Text-to-Image Generative Modelscited by 9
- 2024CBF-LLM: Safe Control for LLM Alignmentcited by 9
- 2024Measuring Copyright Risks of Large Language Model via Partial Information Probingcited by 9
- 2021DP-FP: Differentially Private Forward Propagation for Large Modelscited by 7
- 2025Automating Financial Statement Audits with Large Language Modelscited by 4
Built on by
Citing papers per year: 2025: 1; 2026: 1.
2 papers.
Similar, not cited
Close in wording, no citation either way
- 2024Can Watermarking Large Language Models Prevent Copyrighted Text Generation and Hide Training Data?
- 2024SHIELD: Evaluation and Defense Strategies for Copyright Compliance in LLM Text Generation
- 2024Protecting Copyrighted Material with Unique Identifiers in Large Language Model Training
- 2025Watermarking LLM-Generated Datasets in Downstream Tasks
- 2024Watermarking Large Language Models and the Generated Content: Opportunities and Challenges
- 2025A Survey on Privacy Risks and Protection in Large Language Models
- 2026SoK: Privacy Risks and Mitigations in Retrieval-Augmented Generation Systems
- 2025As If We've Met Before: LLMs Exhibit Certainty in Recognizing Seen Files