# SoundBreak: A Systematic Study of Audio-Only Adversarial Attacks on Trimodal Models Source: [SoundBreak: A Systematic Study of Audio-Only Adversarial Attacks on Trimodal Models]() Collection index: [llms.txt]() Data updated: 2026-01-27T05:00:04.746Z Authors: Aafiya Hussain, Gaurav Srivastava, Alvi Ishmam, Zaber Hakim, Chris Thomas Published: 2026-01-20T18:53:29.000Z ## Abstract Multimodal foundation models that integrate audio, vision, and language achieve strong performance on reasoning and generation tasks, yet their robustness to adversarial manipulation remains poorly understood. We study a realistic and underexplored threat model: untargeted, audio-only adversarial attacks on trimodal audio-video-language models. We analyze six complementary attack objectives that target different stages of multimodal processing, including audio encoder representations, cross-modal attention, hidden states, and output likelihoods. Across three state-of-the-art models and multiple benchmarks, we show that audio-only perturbations can induce severe multimodal failures, achieving up to 96% attack success rate. We further show that attacks can be successful at low perceptual distortions (LPIPS \<= 0.08, SI-SNR \>= 0) and benefit more from extended optimization than increased data scale. Transferability across models and encoders remains limited, while speech recognition systems such as Whisper primarily respond to perturbation magnitude, achieving \>97% attack success under severe distortion. These results expose a previously overlooked single-modality attack surface in multimodal systems and motivate defenses that enforce cross-modal consistency. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Audio Visual Question Answering on AVQA (Robustness Evaluation)](): Audio-Visual Question Answering; AVQA; AVQA Clean Accuracy: 95.6. - [Audio-Visual Question Answering on AVQA (subset 2000 samples)](): Audio-Visual Question Answering; AVQA (subset 2000 samples); ASR Accuracy: 96.03. - [Audio-Visual Question Answering on Music-AVQA 2000 samples](): Audio-Visual Question Answering; Music-AVQA 2000 samples; ASR Rate: 13.8. - [Audio-Visual Scene-aware Dialog on AVSD (val)](): Audio-Visual Scene-Aware Dialog; AVSD (val); ASR (%): 59.48. - [Audio Visual Question Answering on Music-AVQA (Robustness Evaluation)](): Audio-Visual Question Answering; MUSIC-AVQA; Music-AVQA Clean Accuracy: 80.7. ## Original sources and implementation [Original paper]() [PDF]()