Network-based Intrusion Detection has been one of the major focus areas of Cybersecurity since the inception of the modern idea of the "Internet". When Internet traffic was more sparse and operating at specifications such as 10Base2, it may have been possible (albeit unlikely) for analysts and administrators to manually sift through all of the traffic, identify potential threats, and respond accordingly. However, given the exponential increase in Internet traffic and ethernet bandwidth over the past 30 years, this process is simply unthinkable in today's world.
Given the "Big Data" nature of network traffic, it follows that Machine Learning and Artificial Intelligence may be able to assist in identifying and flagging malicious traffic for manual review. In this project, we employ various algorithms to detect simulated malicious network traffic provided by the KDD Cup 1999 Challenge and analyze their performance over different metrics.
H2O - Machine Learning
TensorFlow - Deep Learning
Numpy & Pandas - Data Cleaning / Formatting