Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ERIS: Evolutionary Real-world Interference Scheme for Jailbreaking Audio Large Models

Yibo Zhang, Liang Lin

Abstract

Existing Audio Large Models (ALMs) alignment focuses on clean inputs, neglecting security risks in complex environments. We propose ERIS, a framework transforming real-world interference into a strategically optimized carrier for jailbreaking ALMs. Unlike methods relying on manually designed acoustic patterns, ERIS uses a genetic algorithm to optimize the selection and synthesis of naturalistic signals. Through population initialization, crossover fusion, and probabilistic mutation, it evolves audio fusing malicious instructions with real-world interference. To humans and safety filters, these samples present as natural speech with harmless background noise, yet bypass alignment. Evaluations on multiple ALMs show ERIS significantly outperforms both text and audio jailbreak baselines. Our findings reveal that seemingly innocuous real-world interference can be leveraged to circumvent safety constraints, providing new insights for defensive mechanisms in complex acoustic scenarios.

ERIS: Evolutionary Real-world Interference Scheme for Jailbreaking Audio Large Models

Abstract

Existing Audio Large Models (ALMs) alignment focuses on clean inputs, neglecting security risks in complex environments. We propose ERIS, a framework transforming real-world interference into a strategically optimized carrier for jailbreaking ALMs. Unlike methods relying on manually designed acoustic patterns, ERIS uses a genetic algorithm to optimize the selection and synthesis of naturalistic signals. Through population initialization, crossover fusion, and probabilistic mutation, it evolves audio fusing malicious instructions with real-world interference. To humans and safety filters, these samples present as natural speech with harmless background noise, yet bypass alignment. Evaluations on multiple ALMs show ERIS significantly outperforms both text and audio jailbreak baselines. Our findings reveal that seemingly innocuous real-world interference can be leveraged to circumvent safety constraints, providing new insights for defensive mechanisms in complex acoustic scenarios.

Paper Structure

This paper contains 17 sections, 6 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Safety Alignment in Audio Large Models
  4. Jailbreaking Attacks on Audio Large Models
  5. Impact of Acoustic Interference
  6. Methodology
  7. Problem Formulation
  8. The Design of ERIS
  9. Population Initialization
  10. Crossover Fusion
  11. Probabilistic Mutation
  12. Fitness Evaluation Mechanism
  13. Experiments
  14. Experimental Setup
  15. Main Results
  16. ...and 2 more sections

Figures (2)

  • Figure 1: ERIS transforms a rejected harmful query into an accepted one via evolutionary audio optimization.
  • Figure 2: Top five harmful interferences across different ALMs.