Taylor Unswift: Secured Weight Release for Large Language Models via Taylor Expansion

TL;DR

TaylorMLP introduces a method to secure large language models by transforming weights into Taylor-series parameters, enabling ownership protection and abuse prevention through controlled generation speed, with verified effectiveness across multiple datasets.

Contribution

The paper proposes TaylorMLP, a novel approach that secures LLM ownership and prevents misuse by releasing Taylor-series parameters instead of original weights.

Findings

Induces over 4x increase in latency for token generation.

Maintains token output accuracy matching original LLMs.

Effectively prevents weight reconstruction from downstream data.

Abstract

Ensuring the security of released large language models (LLMs) poses a significant dilemma, as existing mechanisms either compromise ownership rights or raise data privacy concerns. To address this dilemma, we introduce TaylorMLP to protect the ownership of released LLMs and prevent their abuse. Specifically, TaylorMLP preserves the ownership of LLMs by transforming the weights of LLMs into parameters of Taylor-series. Instead of releasing the original weights, developers can release the Taylor-series parameters with users, thereby ensuring the security of LLMs. Moreover, TaylorMLP can prevent abuse of LLMs by adjusting the generation speed. It can induce low-speed token generation for the protected LLMs by increasing the terms in the Taylor-series. This intentional delay helps LLM developers prevent potential large-scale unauthorized uses of their models. Empirical experiments across five datasets and three LLM architectures demonstrate that TaylorMLP induces over 4x increase in latency, producing the tokens precisely matched with original LLMs. Subsequent defensive experiments further confirm that TaylorMLP effectively prevents users from reconstructing the weight values based on downstream datasets.