From the course: Machine Learning for Red Team Hackers by Infosec

Unlock this course with a free trial

Join today to access over 24,700 courses taught by industry experts.

Evading machine learning malware classifiers overview

Evading machine learning malware classifiers overview

From the course: Machine Learning for Red Team Hackers by Infosec

Start my 1-month free trial Buy for my team

Evading machine learning malware classifiers overview

“

(upbeat music) - [Instructor] In the next sequence of lessons, we are going to learn how to evade machine learning malware classifiers. Obviously, the goal here is to test and iteratively improve the machine learning classifier. In particular, we are going to be looking at two classifiers, one is a neural network-based malware classifier called MalConv, about which you can learn more in my other course, Cybersecurity Data Science, where we spend time understanding its architecture and implementing it. The other machine learning classifier is a gradient boosting-based classifier called Ember. Unlike MalConv that relies on pure end-to-end machine learning where the engineering, where the features are engineered automatically, Ember actually has hand-selected features, that is, features that have been engineered by people who know things about malware. By understanding the structure of PE files, we will be able to trick these classifiers into thinking that a malicious sample is actually…

Contents