Scaling Exposes the Trigger: Input-Level Backdoor Detection in Text-to-Image Diffusion Models via Cross-Attention Scaling

TL;DR

This paper introduces SET, a novel input-level backdoor detection method for text-to-image diffusion models that exploits differences in response patterns under cross-attention scaling, outperforming existing defenses especially against stealthy triggers.

Contribution

The work uncovers the Cross-Attention Scaling Response Divergence phenomenon and develops SET, a trigger-agnostic detection framework that learns a benign response space for robust backdoor detection.

Findings

SET outperforms existing methods across diverse attack scenarios.

Achieves 9.1% higher AUROC and 6.5% higher ACC than the best baseline.

Effective against stealthy, implicit-trigger backdoor attacks.

Abstract

Text-to-image (T2I) diffusion models have achieved remarkable success in image synthesis, but their reliance on large-scale data and open ecosystems introduces serious backdoor security risks. Existing defenses, particularly input-level methods, are more practical for deployment but often rely on observable anomalies that become unreliable under stealthy, semantics-preserving trigger designs. As modern backdoor attacks increasingly embed triggers into natural inputs, these methods degrade substantially, raising a critical question: can more stable, implicit, and trigger-agnostic differences between benign and backdoor inputs be exploited for detection? In this work, we address this challenge from an active probing perspective. We introduce controlled scaling perturbations on cross-attention and uncover a novel phenomenon termed Cross-Attention Scaling Response Divergence (CSRD), where benign and backdoor inputs exhibit systematically different response evolution patterns across denoising steps. Building on this insight, we propose SET, an input-level backdoor detection framework that constructs response-offset features under multi-scale perturbations and learns a compact benign response space from a small set of clean samples. Detection is then performed by measuring deviations from this learned space, without requiring prior knowledge of the attack or access to model training. Extensive experiments demonstrate that SET consistently outperforms existing baselines across diverse attack methods, trigger types, and model settings, with particularly strong gains under stealthy implicit-trigger scenarios. Overall, SET improves AUROC by 9.1% and ACC by 6.5% over the best baseline, highlighting its effectiveness and robustness for practical deployment.