# When Backdoors Go Beyond Triggers: Semantic Drift in Diffusion Models Under Encoder Attacks

arXiv: https://arxiv.org/abs/2602.20193 · HTML: https://mlsanity.com/paper/2602.20193/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Robustness, privacy and adversarial aspects, one of 14 directions in Diffusion models.
- Builds on 18 of its 29 references that are in the map.
- First posted on arXiv 21 Feb 2026.

## Placed in

- Diffusion models / [Robustness, privacy and adversarial aspects](https://mlsanity.com/diffusion/robustness-privacy-and-adversarial-aspects/)
- LLM safety and security / [Backdoor and data poisoning attacks](https://mlsanity.com/safety/backdoor-and-data-poisoning-attacks/)

## Builds on (most cited first, 18 of 18)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Momentum Contrast for Unsupervised Visual Representation Learning](https://mlsanity.com/paper/1911.05722/) (2019)
- [Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding](https://mlsanity.com/paper/2205.11487/) (2022)
- [LAION-5B: An open large-scale dataset for training next generation image-text models](https://mlsanity.com/paper/2210.08402/) (2022)
- [CLIPScore: A Reference-free Evaluation Metric for Image Captioning](https://mlsanity.com/paper/2104.08718/) (2021)
- [Improved Baselines with Momentum Contrastive Learning](https://mlsanity.com/paper/2003.04297/) (2020)
- [Reproducible scaling laws for contrastive language-image learning](https://mlsanity.com/paper/2212.07143/) (2022)
- [Intrinsic Dimensionality Explains the Effectiveness of Language Model Fine-Tuning](https://mlsanity.com/paper/2012.13255/) (2020)
- [TIFA: Accurate and Interpretable Text-to-Image Faithfulness Evaluation with Question Answering](https://mlsanity.com/paper/2303.11897/) (2023)
- [Text-to-Image Diffusion Models can be Easily Backdoored through Multimodal Data Poisoning](https://mlsanity.com/paper/2305.04175/) (2023)
- [Nightshade: Prompt-Specific Poisoning Attacks on Text-to-Image Generative Models](https://mlsanity.com/paper/2310.13828/) (2023)
- [VillanDiffusion: A Unified Backdoor Attack Framework for Diffusion Models](https://mlsanity.com/paper/2306.06874/) (2023)
- [ReFACT: Updating Text-to-Image Models by Editing the Text Encoder](https://mlsanity.com/paper/2306.00738/) (2023)
- [Context Matters for Image Descriptions for Accessibility: Challenges for Referenceless Evaluation Metrics](https://mlsanity.com/paper/2205.10646/) (2022)
- [BackdoorDM: A Comprehensive Benchmark for Backdoor Learning on Diffusion Model](https://mlsanity.com/paper/2502.11798/) (2025)
- [Backdooring Self-Supervised Contrastive Learning by Noisy Alignment](https://mlsanity.com/paper/2508.14015/) (2025)

## Abstract (arXiv)

Standard evaluations of backdoor attacks on text-to-image (T2I) models primarily measure trigger activation and visual fidelity. We challenge this paradigm, demonstrating that encoder-side poisoning induces persistent, trigger-free semantic corruption that fundamentally reshapes the representation manifold. We trace this vulnerability to a geometric mechanism: a Jacobian-based analysis reveals that backdoors act as low-rank, target-centered deformations that amplify local sensitivity, causing distortion to propagate coherently across semantic neighborhoods. To rigorously quantify this structural degradation, we introduce SEMAD (Semantic Alignment and Drift), a diagnostic framework that measures both internal embedding drift and downstream functional misalignment. Our findings, validated across diffusion and contrastive paradigms, expose the deep structural risks of encoder poisoning and highlight the necessity of geometric audits beyond simple attack success rates.