ch_11

Chapter 11: Machine Learning Anomaly Detection

This chapter revisits anomaly detection on login attempt data, using machine learning techniques, all while giving you a taste of how the workflow looks in practice.

Content

We will work through five notebooks emulating a real-world scenario using the simulated data in the logs/ directory. Here's a breakdown of the files included in this chapter:

• logs/: Directory containing all simulated log files for the analysis
• user_data/: Directory containing information on the user base used for the simulation (for the simulate.py script to use)
• 0-simulating_the_data.ipynb: Jupyter notebook showing how data was simulated
• 1-EDA_unlabeled_data.ipynb: Jupyter notebook used to perform our EDA of the unlabeled data
• 2-unsupervised_anomaly_detection.ipynb: Jupyter notebook used to test out some unsupervised anomaly detection alogrithms
• 3-EDA_labeled_data.ipynb: Jupyter notebook used to perform our EDA of the labeled data
• 4-supervised_anomaly_detection.ipynb: Jupyter notebook used to build and evaluate supervised anomaly detection models
• 5-online_learning.ipynb: Jupyter notebook used to implement an online learning classifier
• merge_logs.py: Python script for merging the logs of individually simulated months
• run_simulations.sh: Bash script for simulating and merging the log files (this is used to generate the data)
• simulate.py: Python script for simulating the data using the login_attempt_simulator package

The end-of-chapter exercises will use the data in the logs/ directory to explore additional algorithms for machine learning anomaly detection; solutions to these exercises can be found in the repository's solutions/ch_11/ directory.