Deep Learning on Disassembly

Explore deep learning techniques applied to disassembly for malware identification in this 41-minute Black Hat conference talk by Matt Wolff and Andrew Davis. Learn about the pipeline from raw binaries to disassembly data extraction and transformation, and deep learning model training. Discover the effectiveness of these models through presented data and a live demo evaluating them against active malware feeds. Gain insights into topics such as machine learning, input data modalities, feature engineering, logistic regression, recurrent neural networks, and the application of neural networks to disassembly. Understand the potential of treating disassembly as an extension of natural language processing and its implications for malware detection.

Introduction
About Silence
About Malware
Roadmap
Hands
Machine Learning
Input Data Modalities
Labels
Feature Engineering
Logistic Regression
Future Engineering
Automatic Captioning
Failure Mode
Recurrent Neural Network
Paul Graham
Deep Neural Network
Edge Detection
Neural Networks
Disassembly
Takeaways
Questions