Research map / Adversarial robustness and security / Backdoor and poisoning
Handcrafted Backdoors in Deep Neural Networks
Sanghyun Hong, Nicholas Carlini, Alexey Kurakin
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 |
|---|---|---|---|
| handcrafted backdoor instead of poisoning | Backdoor attack on pre-trained models · SVHN CNN, Checkerboard | attack success rate (%): 96 -> 100 | Improved |
| handcrafted backdoor instead of poisoning | Backdoor attack · CIFAR10 CNN, Checkerboard | attack success rate (%): 98 -> 97 | Hurt |
| handcrafted backdoor instead of poisoning with few samples | Backdoor attack · CIFAR10 | attack success rate (%): 100% success with only 50 samples | Improved |
| handcrafted backdoor instead of poisoning under fine-tuning | Backdoor attack · Fine-tuning defense | attack success rate (%): 82 -> 88 (CIFAR10 Checkerboard) | Improved |
| handcrafted backdoor instead of poisoning under fine-pruning | Backdoor attack · Fine-pruning on CIFAR10 and SVHN | attack success rate (%): 96 -> 82 (CIFAR10 Random); 9-27% higher on SVHN | Mixed |
| add random noise or parameter clipping defenses | Handcrafted backdoor · Defense at 5% accuracy drop | attack success rate: over 98% (noise), over 86% (clipping) | No clear effect |
Closest papers by meaning
- Architectural Backdoors in Neural Networks (CVPR 2023 · similarity 0.89)
- BaDExpert: Extracting Backdoor Functionality for Accurate Backdoor Input Detection (ICLR 2024 · similarity 0.87)
- Efficient Backdoor Attacks for Deep Neural Networks in Real-world Scenarios (ICLR 2024 · similarity 0.87)
- A Theoretical Analysis of Backdoor Poisoning Attacks in Convolutional Neural Networks (ICML 2024 · similarity 0.87)
- BackdoorDM: A Comprehensive Benchmark for Backdoor Learning on Diffusion Model (NeurIPS 2025 · similarity 0.87)
- From Toxic to Trustworthy: Using Self-Distillation and Semi-supervised Methods to Refine Neural Networks (AAAI 2024 · similarity 0.87)
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.
