Research map / Adversarial robustness and security / Backdoor and poisoning
Versatile Backdoor Attack With Visible, Semantic, Sample-Specific and Compatible Triggers
Ruotong Wang, Hongrui Chen, Zihao Zhu, Li Liu, Baoyuan Wu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add fine-grained trigger selection (reject low-ISR triggers) (ablation) | VSSC attack · ImageNet-Dogs and FOOD-11, digital | ASR: actual poisoning only 0.91% and 8.34% for low-quality triggers at preset 10% | Improved |
| add quality assessment module (VLM check) to poisoned sample generation (ablation) | VSSC attack · image classification, ResNet-18, digital | ASR: maximum increase of 10.11% in ASR | Improved |
| use advanced diffusion image editing (Null-text Inversion) instead of patching or InstructPix2Pix for trigger insertion (ablation) | VSSC attack · physical scenario classification | ASR | Improved |
| raise poisoning ratio from 5% to 30% (ablation) | VSSC (red flower) · ImageNet-Dogs physical, ResNet-18 | ASR (%): 65.00 -> 89.00 | Improved |
| raise poisoning ratio from 5% to 30% (ablation) | VSSC (strawberry) · FOOD-11 physical, VGG19-BN | ASR (%): 61.00 -> 100.00 | Improved |
| use VSSC semantic trigger instead of BadNets patch on clean accuracy | VSSC · classification digital | C-Acc: less than 2% C-Acc decrease | Improved |
| use VSSC trigger instead of baselines under print-and-recapture distortion | VSSC · ImageNet-Dogs D2P, ResNet-18 | ASR (%): WaNet 76.19 -> VSSC 97.62 | Improved |
| move from digital to digital-to-physical scenario | VSSC · face verification | ASR: decreases by an average of only 5.81% | Hurt |
| add VSSC trigger vs baseline triggers under D2P in face verification | baselines · face verification D2P | ASR: decrease in ASR up to 32.13% | Hurt |
| apply DDE defense to VSSC | VSSC_1 (ResNet-18) · ImageNet-Dogs 5% under defenses | ASR (%): 87.79 | No clear effect |
| apply ABL defense to VSSC_2 versus BadNets | VSSC_2 · ImageNet-Dogs 5% | ASR (%): VSSC_2 73.85 vs BadNets 1.86 | Mixed |
| insert trigger only into compatible classes (collar on animals) (ablation) | VSSC · animals and vehicles dataset | ASR | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- WaNet — WaNet - Imperceptible Warping-based Backdoor Attack (ICLR 2021) · better on 2 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Conditional Backdoor Attack via JPEG Compression (AAAI 2024 · similarity 0.87)
- LOTUS: Evasive and Resilient Backdoor Attacks through Sub-Partitioning (CVPR 2024 · similarity 0.87)
- A Dual Stealthy Backdoor: From Both Spatial and Frequency Perspectives (AAAI 2024 · similarity 0.87)
- Temporal-Distributed Backdoor Attack against Video Based Action Recognition (AAAI 2024 · similarity 0.87)
- Revisiting Backdoor Attacks against Large Vision-Language Models from Domain Shift (CVPR 2025 · similarity 0.87)
- A Set of Generalized Components to Achieve Effective Poison-only Clean-label Backdoor Attacks with Collaborative Sample Selection and Triggers (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
