[2509.11128] ERIS: Evolutionary Real-world Interference Scheme for Jailbreaking Audio Large Models

Computer Science > Sound arXiv:2509.11128 (cs) [Submitted on 14 Sep 2025 (v1), last revised 2 Mar 2026 (this version, v2)] Title:ERIS: Evolutionary Real-world Interference Scheme for Jailbreaking Audio Large Models Authors:Yibo Zhang, Liang Lin View a PDF of the paper titled ERIS: Evolutionary Real-world Interference Scheme for Jailbreaking Audio Large Models, by Yibo Zhang and 1 other authors View PDF HTML (experimental) 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. Subjects: Sound (cs.SD); Artificial Intelligence (cs.AI) Cite as: ar...