DeepLocker - Concealing Targeted Attacks with AI Locksmithing

Explore the concept of DeepLocker, a novel class of highly targeted and evasive attacks powered by artificial intelligence, in this 53-minute Black Hat conference talk. Delve into the mechanisms and implications of malicious AI use in cybersecurity, understanding how cybercriminals weaponize AI and the importance of staying ahead of these threats. Learn about recent examples of AI in attacks, the evolution of evasive malware, and the inner workings of DeepLocker, including deep neural networks, target attributes, key generation, and concealing techniques. Discover the challenges posed by AI-powered attacks, the concept of AI locksmithing, and the computational costs involved. Gain insights into attribute targeting and face detection as potential attack vectors, equipping yourself with knowledge to deploy appropriate defenses against these sophisticated threats.

Introduction
Recent examples of AI in attacks
Evolution of evasive malware
DeepLocker
Deep Neural Networks
Target Attributes
Overview
Key Generation
Analysis
Concealing
The Wakeup Call
The Problem
What is Lockpicking
Summary
Questions
Computational Cost
Attribute Targeting
Face Detection