SWaRL: Safeguard Code Watermarking via Reinforcement Learning

2026-01-05

arXiv.org

AI Summary

Existing watermarking methods for large code models often compromise functional correctness by introducing syntactic or semantic errors, making it difficult to simultaneously ensure watermark detectability and code usability. This work proposes a reinforcement learning–based co-training framework that, for the first time, integrates compiler feedback with a confidential verifier: compiler signals guarantee 100% functional correctness of generated code, while the verifier provides rewards to maintain high watermark detection accuracy. By incorporating LoRA (Low-Rank Adaptation), the approach enables transferable embedding of watermark information during model updates, significantly enhancing robustness against code refactoring and adversarial transformations with minimal computational overhead.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Large Vision Models
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

We present SWaRL, a robust and fidelity-preserving watermarking framework designed to protect the intellectual property of code LLM owners by embedding unique and verifiable signatures in the generated output. Existing approaches rely on manually crafted transformation rules to preserve watermarked code functionality or manipulate token-generation probabilities at inference time, which are prone to compilation errors. To address these challenges, SWaRL employs a reinforcement learning-based co-training framework that uses compiler feedback for functional correctness and a jointly trained confidential verifier as a reward signal to maintain watermark detectability. Furthermore, SWaRL employs low-rank adaptation (LoRA) during fine-tuning, allowing the learned watermark information to be transferable across model updates. Extensive experiments show that SWaRL achieves higher watermark detection accuracy compared to prior methods while fully maintaining watermarked code functionality. The LoRA-based signature embedding steers the base model to generate and solve code in a watermark-specific manner without significant computational overhead. Moreover, SWaRL exhibits strong resilience against refactoring and adversarial transformation attacks.

Problem

• code watermarking
• intellectual property protection
• functional correctness
• compiler errors
• watermark detectability

Innovation

• Reinforcement Learning
• Code Watermarking
• Low-Rank Adaptation (LoRA)
• Compiler Feedback
• Intellectual Property Protection