Face-Adversarial-Attack-and-Defense

Practice

对抗攻击介绍    Facenet    FaceSwap    Adversarial-Attack    AdvBox    IBM-ART    GUI   

Aim

1. Familiar with the Facenet opensource tool.
2. Familiar with the knowledge about adversarial attack.
3. Add the adversary attack to the Facenet to generate ideal adversary example.
4. Evaluation the attack effect of adversary face image example.

References

1. Facenet

2. A Chinese Facenet tutorial

3. Adversarial-Face-Attack

4. Cleverhans : A software library provides standardized reference implementations of adversarial example && training

5. CleverHas Blog

6. Pytorch tutorial about adversarial example generation and FGSM attack

7. CW自己的算法

8. The links to the MCS 2018 Github repositories, the other adversary attack algorithms:

   • MCS 2nd place

     • repository https://github.com/snakers4/msc-2018-final

     • blog page https://spark-in.me/post/playing-with-mcs2018-adversarial-attacks

   • MCS 14th place

     • repository https://github.com/miltonbd/mcs_2018_adversarial_attack

Some Open Source

• Synced

• Adversarial-Face-Attack

• Adversarial-Occlussion-aware-Face-Detection

• Fake Face Detect

• face-swap topic

• FaceForensics

• Learn-Convolutional-Neural-Network-for-Face-Anti-Spoofing

• handong1587.github.io

• adversarial-machine-learning topic

Documentation Examples

• artemis

• Near and long-term directions for adversarial AI in cybersecurity

• 保护AI模型免受音频对抗攻击

• 破解深度学习模型：您甚至不需要博士学位或GPU

• 对抗攻击（Adversarial attacks）的常用术语

• AI对抗攻击

• 一张贴纸欺骗AI，对抗性补丁让人类隐身

• 多伦多大学反人脸识别，身份欺骗成功率达99.5%

• 对抗样本：知其然，知其所以然

• 逃避对机器学习的攻击

• 对抗攻击论文

• MCS2018_挑战

• 对抗鲁棒性