Published December 15, 2025 | Version v1

The Prompt Stealing Fallacy: Rethinking Metrics, Attacks, and Defenses

  • 1. ROR icon Swinburne University of Technology

Description

The Prompt Stealing Fallacy: Rethinking Metrics, Attacks, and Defenses

1. Overview

This repository contains the implementation of `The Prompt Stealing Fallacy: Rethinking Metrics, Attacks, and Defenses`. We present one novel attack method and two defense strategies to address the vulnerability of prompt stealing in T2I Models.
 
2. Attack Method

2.1 Attack
Located in `./pthieft_v01/attack/`

Our attack leverages the MS-SWIFT framework to perform prompt stealing attacks on T2I models. The attack method is designed to extract high-contribution prompts.
 
3. Defense Methods

3.1  PSA Defense Framework
Located in `./pthieft_v01/defense`
See `./pthieft_v01/defense/README.md` for defense setup and run instructions.

Files

pthieft_v01.zip

Files (91.8 MB)

Name Size Download all
md5:51e62f2d72b69c1796c973a5b90d3d83
91.8 MB Preview Download

Additional details

Software

Programming language
Python
Development Status
Active