Research map / Adversarial robustness and security / Attacks on text and graphs
Topological Detection of Trojaned Neural Networks
Songzhu Zheng, Yikai Zhang, Hubert Wagner, Mayank Goswami, Chao Chen
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 |
|---|---|---|---|
| Use topological features instead of correlation-matrix features | Trojan detector · MNIST+LeNet5 | ACC: 0.59±0.10 -> 0.85±0.07 | Improved |
| Use topological features instead of correlation-matrix features | Trojan detector · MNIST+LeNet5 | AUC: 0.62±0.10 -> 0.89±0.04 | Improved |
| Use topological features instead of ULP | Trojan detector · MNIST+ResNet18 | ACC: 0.71±0.14 -> 0.87±0.09 | Improved |
| Use topological features instead of ULP | Trojan detector · MNIST+ResNet18 | AUC: 0.71±0.14 -> 0.97±0.02 | Improved |
| Use topological features instead of correlation-matrix features | Trojan detector · CIFAR10+ResNet18 | ACC: 0.72±0.07 -> 0.93±0.06 | Improved |
| Use topological features instead of correlation-matrix features | Trojan detector · CIFAR10+ResNet18 | AUC: 0.81±0.08 -> 0.97±0.02 | Improved |
| Use topological features instead of DFTND | Trojan detector · CIFAR10+Densenet121 | ACC: 0.59±0.07 -> 0.84±0.04 | Improved |
| Use topological features instead of correlation-matrix features | Trojan detector · CIFAR10+Densenet121 | AUC: 0.66±0.07 -> 0.93±0.03 | Improved |
| Use topological features instead of ULP | Trojan detector · Round1-ResNet | ACC: 0.63±0.00 -> 0.77±0.04 | Improved |
| Use topological features instead of ULP | Trojan detector · Round1-ResNet | AUC: 0.62±0.03 -> 0.87±0.03 | Improved |
| Use topological features instead of ULP | Trojan detector · Round1-DenseNet | ACC: 0.63±0.06 -> 0.62±0.04 | Hurt |
| Use topological features instead of ULP | Trojan detector · Round1-DenseNet | AUC: 0.63±0.06 -> 0.69±0.04 | Improved |
Closest papers by meaning
- Practical Detection of Trojan Neural Networks: Data-Limited and Data-Free Cases (ECCV 2020 · similarity 0.89)
- Trigger Hunting with a Topological Prior for Trojan Detection (ICLR 2022 · similarity 0.88)
- Scanning Trojaned Models Using Out-of-Distribution Samples (NeurIPS 2024 · similarity 0.84)
- Deep-TROJ: An Inference Stage Trojan Insertion Algorithm through Efficient Weight Replacement Attack (CVPR 2024 · similarity 0.83)
- Concept-ROT: Poisoning Concepts in Large Language Models with Model Editing (ICLR 2025 · similarity 0.83)
- When Witnesses Defend: A Witness Graph Topological Layer for Adversarial Graph Learning (AAAI 2025 · similarity 0.83)
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.
