Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.08280 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917474501394432 |
|---|---|
| author | Bowen, Lu Tang, Xinyu Low, Yin Yin Leong, Shu-Min |
| author_facet | Bowen, Lu Tang, Xinyu Low, Yin Yin Leong, Shu-Min |
| contents | Preserving model fidelity is essential for stealthy text-to-image (T2I) backdoor attacks. Existing methods such as Learning without Forgetting (LwF) rely on output-based distillation, which provides limited regularization. We introduce Elastic Weight Consolidation (EWC) as a parameter-based alternative for preserving fidelity in backdoor learning. While stronger in principle, we show that standard static EWC with a fixed regularization weight lambda and mean-squared utility loss creates an artificial trade-off between attack success rate (ASR) and fidelity, particularly degrading performance on weak triggers. To address this, we propose Cosine-Aware Adaptive EWC, which dynamically adjusts EWC regularization using a cosine-based semantic utility and adaptive scheduling. This approach transforms EWC from a fixed penalty into a context-sensitive constraint, maintaining high ASR while preserving model fidelity. Experiments demonstrate improved ASR-fidelity balance and enhanced robustness on out-of-domain (OOD) datasets compared to existing baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_08280 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Beyond the False Trade-off: Adaptive EWC for Stealthy and Generalizable T2I Backdoors Bowen, Lu Tang, Xinyu Low, Yin Yin Leong, Shu-Min Machine Learning Artificial Intelligence Preserving model fidelity is essential for stealthy text-to-image (T2I) backdoor attacks. Existing methods such as Learning without Forgetting (LwF) rely on output-based distillation, which provides limited regularization. We introduce Elastic Weight Consolidation (EWC) as a parameter-based alternative for preserving fidelity in backdoor learning. While stronger in principle, we show that standard static EWC with a fixed regularization weight lambda and mean-squared utility loss creates an artificial trade-off between attack success rate (ASR) and fidelity, particularly degrading performance on weak triggers. To address this, we propose Cosine-Aware Adaptive EWC, which dynamically adjusts EWC regularization using a cosine-based semantic utility and adaptive scheduling. This approach transforms EWC from a fixed penalty into a context-sensitive constraint, maintaining high ASR while preserving model fidelity. Experiments demonstrate improved ASR-fidelity balance and enhanced robustness on out-of-domain (OOD) datasets compared to existing baselines. |
| title | Beyond the False Trade-off: Adaptive EWC for Stealthy and Generalizable T2I Backdoors |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2605.08280 |