From Essence to Defense: Adaptive Semantic-aware Watermarking for Embedding-as-a-Service Copyright Protection
Not on the map. Why?
Abstract (arXiv)
Benefiting from the superior capabilities of large language models in natural language understanding and generation, Embeddings-as-a-Service (EaaS) has emerged as a successful commercial paradigm on the web platform. However, prior studies have revealed that EaaS is vulnerable to imitation attacks. Existing methods protect the intellectual property of EaaS through watermarking techniques, but they all ignore the most important properties of embedding: semantics, resulting in limited harmlessness and stealthiness. To this end, we propose SemMark, a novel semantic-based watermarking paradigm for EaaS copyright protection. SemMark employs locality-sensitive hashing to partition the semantic space and inject semantic-aware watermarks into specific regions, ensuring that the watermark signals remain imperceptible and diverse. In addition, we introduce the adaptive watermark weight mechanism based on the local outlier factor to preserve the original embedding distribution. Furthermore, we propose Detect-Sampling and Dimensionality-Reduction attacks and construct four scenarios to evaluate the watermarking method. Extensive experiments are conducted on four popular NLP datasets, and SemMark achieves superior verifiability, diversity, stealthiness, and harmlessness.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 26 of its 45 references that are in the map. First posted on arXiv 18 Dec 2025.
Builds on
26 of 45 references.
References · 26
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
- 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
- 2013Efficient Estimation of Word Representations in Vector Spacecited by 4,346
- 2016Pointer Sentinel Mixture Modelscited by 2,992
- 2017A Broad-Coverage Challenge Corpus for Sentence Understanding through Inferencecited by 2,665
- 2015Character-level Convolutional Networks for Text Classificationcited by 2,293
- 2025Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Modelscited by 1,175
- 2023REPLUG: Retrieval-Augmented Black-Box Language Modelscited by 786
- 2023In-Context Retrieval-Augmented Language Modelscited by 694
- 2022MTEB: Massive Text Embedding Benchmarkcited by 623
- 2016Stealing Machine Learning Models via Prediction APIscited by 602
- 2018What you can cram into a single vector: Probing sentence embeddings for linguistic propertiescited by 452
14 more
- 2018Knockoff Nets: Stealing Functionality of Black-Box Modelscited by 285
- 2024jina-embeddings-v3: Multilingual Embeddings With Task LoRAcited by 138
- 2019Thieves on Sesame Street! Model Extraction of BERT-based APIscited by 136
- 2020Imitation Attacks and Defenses for Black-box Machine Translation Systemscited by 78
- 2023Are You Copying My Model? Protecting the Copyright of Large Language Models for EaaS via Backdoor Watermarkcited by 61
- 2022Towards Data-Free Model Stealing in a Hard Label Settingcited by 57
- 2025A Survey on Model Extraction Attacks and Defenses for Large Language Modelscited by 18
- 2023Dual Student Networks for Data-Free Model Stealingcited by 17
- 2021Student Surpasses Teacher: Imitation Attack for Black-Box NLP APIscited by 16
- 2023Evaluating Embedding APIs for Information Retrievalcited by 11
- 2024WARDEN: Multi-Directional Backdoor Watermarks for Embedding-as-a-Service Copyright Protectioncited by 9
- 2013The Expressive Power of Word Embeddingscited by 7
- 2025Medical Multimodal Model Stealing Attacks via Adversarial Domain Alignmentcited by 5
- 2024Robust and Minimally Invasive Watermarking for EaaScited by 4
Built on by
Citing papers per year: 2026: 2.
2 papers.