Machine Learning Vulnerabilities - CodingNomads Brown Bag Session #2 with Gilad Gressel

CodingNomads Brown Bag Session #2 with Gilad Gressel on Machine Learning and its inherent - somewhat confusing and curious - security vulnerabilities. Join the conversation on CodingNomads Forum: http://forum.codingnomads.co/. ## Adversarial Attacks - Demo's ​ [Adversarial Patch](https://www.youtube.com/watch?v=i1sp4X57TL4) Toaster ​ [Adversarial Patch](https://www.youtube.com/watch?v=MIbFvK2S9g8) vs YOLO video recognition ​ [Adversarial Patch for Object Detection](https://www.youtube.com/watch?v=WXnQjbZ1e7Y) Same idea but now it suppresses all detections on the screen! ​ [Download the paper](https://arxiv.org/pdf/1712.09665v2.pdf) and print out the patch yourself (on page 5) ​ [Turtle is a gun](https://www.theverge.com/2017/11/2/16597276/google-ai-image-attacks-adversarial-turtle-rifle-3d-printed) ​ [Adversarial Glasses](https://www.theverge.com/2016/11/3/13507542/facial-recognition-glasses-trick-impersonate-fool) to become a celebrity ​ ## Good blog / Primers ​ [Good Overview](https://medium.com/element-ai-research-lab/tricking-a-machine-into-thinking-youre-milla-jovovich-b19bf322d55c) by element AI ​ [Breaking Things is Easy -- CleverHans Blog](http://www.cleverhans.io/security/privacy/ml/2016/12/16/breaking-things-is-easy.html) ​ [In Depth Tutorial](https://adversarial-ml-tutorial.org/) by Alex Madry ​ ## Libraries ​ [Adversarial Robustness Toolbox ART](https://github.com/Trusted-AI/adversarial-robustness-toolbox) - IBM ​ [TextAttack](https://github.com/QData/TextAttack) Attacking NLP models ​ [SecML](https://secml.gitlab.io/) ​ [Cleverhans](https://github.com/cleverhans-lab/cleverhans) ​ [Counterfit](https://github.com/Azure/counterfit/) - this one is for secure risk assessment (it uses ART and TextAttack)