SafeSeal is an advanced watermarking technology designed to securely embed identifiable marks in large language model outputs while preserving content quality and robustness against removal attempts.
Background :
The rapid deployment and widespread use of large language models (LLMs) have raised significant concerns about intellectual property protection and unauthorized usage. Conventional watermarking methods often degrade text quality or are vulnerable to attacks that remove or obscure the watermark, undermining security. These challenges necessitated the development of a more reliable and content-preserving watermarking solution to protect proprietary LLM outputs from theft and misuse.
Technology Overview :
SafeSeal introduces a novel watermarking mechanism that strategically embeds watermarks within LLM-generated texts without compromising semantic integrity or readability. It achieves this by leveraging named entity recognition techniques to preserve key entities in the content, ensuring that important information remains unaltered. Meanwhile, it replaces certain linguistic elements with carefully selected, context-aware synonyms to complicate attempts to remove the watermark, therefore enhancing security against adversaries aiming to steal or manipulate model outputs. One of SafeSeal's core innovations is its balance between watermark detectability and content preservation. Through extensive evaluation, SafeSeal attains a BERTScore of 0.981, reflecting very high semantic similarity between original and watermarked texts, and an entity similarity score of 0.962, indicating minimal distortion of named entities. Additionally, the watermark detection rate reaches 95.10%, demonstrating the technology’s reliability in asserting the authenticity of the outputs. By integrating these methods, SafeSeal not only protects intellectual property by embedding verifiable marks but also maintains the utility and quality of LLM-generated content, making it a valuable tool for developers and organizations deploying large language models in sensitive or proprietary applications.
Advantages :
• High semantic fidelity: Maintains meaning and coherence with a BERTScore of 0.981.
• Entity preservation: Protects named entities to avoid content distortion, achieving an entity similarity of 0.962.
• Robust watermark detectability: Achieves a detection rate of 95.10%, ensuring reliable verification.
• Resistance to removal attacks: Uses context-aware synonym replacement with uniform distribution to thwart adversarial attempts.
• Preserves text utility: Ensures that watermark embedding does not degrade the model’s output quality.
• Enhances intellectual property protection in LLM deployments, reducing risks of unauthorized use or model stealing.
Applications :
• Protecting proprietary content generated by commercial large language models.
• Verifying the authenticity of AI-generated documents in industries requiring content integrity.
• Deterring unauthorized copying or redistribution of LLM outputs through embedded watermarks.
• Supporting legal and compliance frameworks by providing traceable and verifiable AI content.
• Enhancing security measures in AI service deployments against manipulation and intellectual property theft.
Intellectual Property Summary : Patent Pending
Stage of Development : TRL = 3
Licensing Status : This technology is available for licensing.
About the Research Foundation for the State University of New York:
As the nation's largest research foundation supporting the nation's largest public university system, The Research Foundation for SUNY powers research and innovation to address today's most pressing problems and shape a better future for generations to come. The Research Foundations supports SUNY researchers leading the way globally in AI for the public good, quantum technologies, next-generation semiconductors, biotech and medicine, energy and climate solutions, and more. The Research Foundation for SUNY is a private, nonprofit educational corporation that is tax-exempt under Internal Revenue Code (IRC) Section 501(c)(3). To learn more, please visit us online at rfsuny.org .
About the State University of New York
The State University of New York is the largest comprehensive system of higher education in the United States, and more than 95 percent of all New Yorkers live within 30 miles of any one of SUNY’s 64 colleges and universities. Across the system, SUNY has four academic health centers, five hospitals, four medical schools, two dental schools, a law school, the country’s oldest school of maritime, the state's only college of optometry, 12 Educational Opportunity Centers, over 30 ATTAIN digital literacy labs, and manages one US Department of Energy National Laboratory. In total, SUNY serves about 1.7 million students across its portfolio of credit- and non-credit-bearing courses and programs, continuing education, and community outreach programs. SUNY oversees nearly a quarter of academic research in New York. Research expenditures system-wide are nearly $1.5 billion in fiscal year 2025, including significant contributions from students and faculty. There are more than three million SUNY alumni worldwide, and annually one in three New Yorkers who earn a college degree is a SUNY alum. To learn more about how SUNY creates opportunities, visit suny.edu .