Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models

2026-05-18

AI Summary

This work reveals, for the first time, the presence of cross-modal backdoor vulnerabilities in unified autoregressive multimodal generative models, wherein an adversary can exploit a common textual trigger to simultaneously manipulate both generated images and text. To address this threat, the authors propose ToBAC, the first targeted attack framework designed specifically for such models. ToBAC leverages either data poisoning or model implantation to inject malicious behavior that propagates covertly through shared parameters and a unified multimodal vocabulary. Experimental results demonstrate that with access to the target model, the attack achieves a 55% success rate on Liquid; even without model access, data poisoning alone yields an average attack success rate of 63.1% on JanusPro, highlighting the significant security risks inherent in current multimodal architectures.

Abstract

Unified autoregressive models (UAMs) are transformer models that generate text as well as image tokens within a single autoregressive pass. Shared parameters and a multimodal vocabulary simplify the training pipeline and facilitate flexible multimodal generation, yet might introduce new vulnerabilities. In particular, we are the first to show that this unified architecture enables multimodal backdoor attacks, where a trigger can propagate malicious effects across multiple output modalities. Specifically, we present the Token by Token Backdoor Attack (ToBAC), the first backdoor attack targeting UAMs, exploring both data-based and model-based poisoning strategies. We demonstrate that innocuous characters or even common words can be transformed into triggers that elicit harmful behavior in autoregressive image generation. ToBAC can jointly manipulate visual outputs and accompanying text, increasing the perceived authenticity of fabricated content. With model access, ToBAC enables attacks on the unified Liquid model in which a subtle word (e.g., ``cool'') induces modality-aligned brand promotion or ideological influence in 55% of generations. Without model access, ToBAC can be induced through data poisoning, achieving an average success rate of 63.1% against JanusPro.

Problem

• backdoor attack
• unified autoregressive models
• multimodal generation
• trigger propagation
• model poisoning

Innovation

• unified autoregressive models
• multimodal backdoor attack
• ToBAC
• cross-modal manipulation
• data poisoning