Transferable Adversarial Perturbations

AI Security AI Village10 minOpen on YouTube ↗

Creating transferable adversarial perturbations that reliably fool multiple disparate machine learning architectures.

Summary

AI adversarial attacks: Transferable Adversarial Perturbations. At 3:00-4:00, attack methodology introduced. Mid-presentation demonstrates practical exploitation.

(AI summary — Claude Haiku)

Related talks

AI Village·33 min

DEF CON
AI Security
Date: Aug 2018
Location: Las Vegas

Practical Adversarial Attacks

Moustafa Alzantot · Yash Sharma

Practical black-box and gradient-free adversarial attack methodologies against machine learning systems.

adversarial-machine-learningblack-box-attacksevasion+1

AI Village·19 min

DEF CON
AI Security
Date: Aug 2018
Location: Las Vegas

Recent Progress in Adversarial Deep Learning

Guo · Cuevas

An overview of recent progress and developments in adversarial deep learning attacks and defenses.

adversarial-machine-learningdeep-learningevasion+1

AI Village·22 min

DEF CON
AI Security
Date: Aug 2018
Location: Las Vegas

Targeted Adversarial Examples for Black Box Audio

Rohan Taori · Ameya Kamsetty

Generating targeted adversarial perturbation examples against black-box automated speech recognition systems.

adversarial-machine-learningaudio-processingspeech-recognition+1