EmMark: Robust Watermarks for IP Protection of Embedded Quantized Large Language Models

arXiv:2402.17938v1 Announce Type: new
Abstract: This paper introduces EmMark,a novel watermarking framework for protecting the intellectual property (IP) of embedded large language models deployed on resource-constrained edge devices. To address the IP theft risks posed by malicious end-users, EmMark enables proprietors to authenticate ownership by querying the watermarked model weights and matching the inserted signatures. EmMark's novelty lies in its strategic watermark weight parameters selection, nsuring robustness and maintaining model quality. Extensive proof-of-concept evaluations of models from OPT and LLaMA-2 …

address arxiv authenticate cs.cl cs.cr devices edge edge devices embedded end framework intellectual property ip protection ip theft language language models large malicious novel ownership property protecting protection resource risks theft watermarking watermarks