PragLocker: Protecting Agent Intellectual Property in Untrusted Deployments via Non-Portable Prompts
Qinfeng Li, Yuntai Bao, Jianghui Hu, Wenqi Zhang, Jintao Chen + 3 more
TLDR
PragLocker protects valuable LLM agent prompts from unauthorized reuse by making them non-portable to other LLMs, securing intellectual property.
Key contributions
- PragLocker protects valuable LLM agent prompts from unauthorized reuse in untrusted environments.
- It constructs function-preserving, obfuscated prompts anchored with code symbols and injected noise.
- Prompts become non-portable, working only on the target LLM while maintaining original performance.
- Demonstrates robustness against adaptive attackers and significantly reduces cross-LLM portability.
Why it matters
LLM agent prompts are valuable intellectual property, but current deployments lack protection against unauthorized copying and reuse. PragLocker offers a crucial solution by ensuring prompts only work on their intended LLM, safeguarding economic value. This innovation is vital for secure and trusted LLM agent development.
Original Abstract
LLM agents rely on prompts to implement task-specific capabilities based on foundation LLMs, making agent prompts valuable intellectual property. However, in untrusted deployments, adversaries can copy and reuse these prompts with other proprietary LLMs, causing economic losses. To protect these prompts, we identify four key challenges: proactivity, runtime protection, usability, and non-portability that existing approaches fail to address. We present PragLocker, a prompt protection scheme that satisfies these requirements. PragLocker constructs function-preserving obfuscated prompts by anchoring semantics with code symbols and then using target-model feedback to inject noise, yielding prompts that only work on the target LLM. Experiments across multiple agent systems, datasets, and foundation LLMs show that PragLocker substantially reduces cross-LLM portability, maintains target performance, and remains robust against adaptive attackers.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.