Ghost in the Transformer: Detecting Model Reuse with Invariant Spectral Signatures

Large Language Models (LLMs) are widely adopted, but their high training cost leads many developers to fine-tune existing open-source models. While most adhere to opensource licenses, some falsely claim original training despite clear derivation from public models, raising pressing concerns about intellectual property protection and the need to verify model provenance. In this paper, we propose Ghost-Spec, a lightweight yet effective method for verifying LLM lineage without access to training data or modification of model behavior. Our approach constructs compact and robust fingerprints by applying singular value decomposition (SVD) to invariant products of internal attention weight matrices. Unlike watermarking or output-based methods, Ghost-Spec is fully data-free, non-invasive, and computationally efficient. Extensive experiments show it is robust to fine-tuning, pruning, expansion, and adversarial transformations, reliably tracing lineage with minimal overhead. By offering a practical solution for model verification, our method contributes to intellectual property protection and fosters a transparent, trustworthy LLM ecosystem.