4 February 2026 in ACM Computing Surveys
State-of-the-art watermarking and fingerprinting techniques for Large Language Models (LLMs) are explored, with our analysis spanning a wide array of methodologies designed to protect the intellectual property of LLMs. The review of watermarking techniques is based on embedding watermarks during the training, logits generation, and token sampling phases. Meanwhile, we investigate the application of watermarking technology in multimodal LLMs and potential attacks on watermarks. Moreover, our examination of fingerprinting techniques revealed the ingenuity behind methods used to identify LLMs. We discussed the development of fingerprints based on model behavior and using deep learning models to learn thresholds for fingerprint comparison. Our survey has underscored the importance of advancing security measures for LLMs, especially in light of the increasing sophistication of adversarial attacks. As LLMs continue to play a pivotal role in advancing AI technologies, developing and refining security measures that safeguard their intellectual property and ensure their ethical deployment is imperative.
3 March 2025 in Computer Standards & Interfaces
12 August 2026 in International Journal of Educational Technology in Higher Education
Online Education in Higher Education is rapidly evolving through the integration of Large Language Model (LLM)-powered intelligent systems, which enable personalized tutoring, dynamic content generation, and automated assessment. However, the widespread adoption of LLMs in education is hampered due to their inherent limitations, including susceptibility to hallucinations, insufficient domain-specific knowledge validation, and output inconsistency. These deficiencies can lead to misleading or erroneous content, potentially causing significant negative learning outcomes. A core challenge lies in ensuring that such errors are immutably logged and traceable, thereby establishing a mechanism for accountability among the entities deploying these LLM services. To address these challenges, this paper proposes a novel framework that Integrates LLM with consortium blockchain for personalized and verifiable online education. Our design features a synergistic architecture in which LLM based services provide the intelligent educational interface, while a permissioned consortium blockchain serves as a secure and tamper proof ledger. This blockchain infrastructure records critical educational transactions ranging from learning process data and academic credentials to the outputs generated by the LLMs. This integration not only secures academic credentials but also establishes a fully auditable trail, making it possible to trace responsibility for educational deficiencies caused by AI errors. Collectively, this work demonstrates a robust and accountable framework for leveraging LLMs in education, effectively mitigating the risks of AI inaccuracies through the verifiable and immutable nature of consortium blockchain.
5 February 2026 in Computer Standards & Interfaces
9 March 2026 in IEEE Transactions on Dependable and Secure Computing
The widespread deployment of large language models (LLMs) in downstream applications has increased the demand for high-quality system prompts, which have become valuable intellectual assets in the commercial prompt marketplace. Recent studies have demonstrated that system prompts are vulnerable to prompt stealing attacks, where adversaries can extract system prompts from LLM applications by crafting adversarial queries, thereby compromising developers' intellectual property and undermining existing business models. However, prior attack methods suffer from critical limitations including gradient dependency and poor scalability, severely restricting their practical applicability. To address these limitations, we present AGFPS, an automated gradient-free framework that leverages evolutionary optimization to systematically steal prompts. Our approach formulates prompt stealing as a discrete optimization problem, where adversarial queries are modeled as individuals in an evolving population. These individuals are optimized through elite retention, selection, adaptive crossover, and mutation operations. To mitigate local optima convergence, we introduce a progressive fitness evaluation strategy based on adaptive sequence fragmentation that exploits LLMs' autoregressive properties. Comprehensive evaluations across multiple benchmark datasets and mainstream LLMs demonstrates that AGFPS achieves a 95.2% exact system prompt stealing success rate, significantly outperforming manual baselines and surpassing gradient-based methods in 80.6% of scenarios. The generated adversarial queries exhibit remarkable transferability across heterogeneous models and diverse datasets, while maintaining robustness against various defense mechanisms. Our work exposes critical vulnerabilities in current LLM deployment practices and underscores the urgent need for enhanced security measures in LLM applications.