Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networks
论文简介 在这项工作中，我们研究了一种新的攻击类型，称为干净标签攻击，攻击者注入的训练示例被认证机构清晰地标记，而不是被攻击者自己恶意地贴上标签。我们的策略假设攻击者不了解训练数据，而是了解模型及其参数。攻击者的目标是当网络在包含中毒实例的增强数据集上进行重新训练后，使重新训练的网络将一个特定测试实例从一个类错误地分类为她选择的另一个类。除了目标的预期预测错误之外，受害的分类器的性能下降并不明显。
相关文章
- 1. 【翻译】Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networks
- 2. Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networks 论文阅读、复现及思考
- 3. Cascade-based attacks on complex networks
- 4. 【翻译】Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks
- 5. [S&P 2019翻译]Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks
- 6. 读书笔记17：Adversarial Attacks on Neural Networks for Graph Data
- 7. 【生成对抗样本】Simple Black-Box Adversarial Attacks on Deep Neural Networks
- 8. Stronger Data Poisoning Attacks Break Data Sanitization Defenses
- 9. Paper Notes: A Comprehensive Survey on Graph Neural Networks
- 10. [娜璋带你读论文] (02) SP2019-Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks
- 更多相关文章...
- • Docker Compose - Docker教程
- • ionic 手势事件 - ionic 教程
- • RxJava操作符（一）Creating Observables
- • PHP Ajax 跨域问题最佳解决方案
相关标签/搜索
每日一句
-
每一个你不满意的现在，都有一个你没有努力的曾经。
欢迎关注本站公众号,获取更多信息
相关文章
- 1. 【翻译】Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networks
- 2. Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networks 论文阅读、复现及思考
- 3. Cascade-based attacks on complex networks
- 4. 【翻译】Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks
- 5. [S&P 2019翻译]Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks
- 6. 读书笔记17：Adversarial Attacks on Neural Networks for Graph Data
- 7. 【生成对抗样本】Simple Black-Box Adversarial Attacks on Deep Neural Networks
- 8. Stronger Data Poisoning Attacks Break Data Sanitization Defenses
- 9. Paper Notes: A Comprehensive Survey on Graph Neural Networks
- 10. [娜璋带你读论文] (02) SP2019-Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks