Outcomes
- Translate behavioral traces into model-ready features
- Compare classical and modern ML approaches for bot classification
Prerequisites
- Sections 1-5 completed
- Comfort with basic statistics and model evaluation
Section Breakdown
Lecture 29
Introduction to Behavioral Biometrics
Explain which behavioral signals are stable enough to matter and which quickly collapse under real-world noise.
Lecture 30
Collecting Behavioral Data
Design an event collection strategy that balances analytic value, privacy, and operational overhead.
Lecture 31
Feature Engineering — From Raw Signals to Model Inputs
Move from clickstream fragments and timing sequences into features that are robust enough for modeling.
Lecture 32
ML Models for Bot Classification
Compare simple baselines, tree-based models, and deeper approaches in the context of bot detection.
Lecture 33
Transformer & LLM Models for Behavioral Analysis
Assess sequence models and LLM-adjacent approaches for representing and explaining complex user flows.
Lecture 34
Model Explainability — SHAP Values & Feature Importance
Keep model outputs interpretable enough for tuning, incident response, and false-positive review.
Lecture 35
Emerging Evasion — AI-Generated Behavior & CAPTCHA Solving Services
Review how current adversaries are synthesizing behavior and outsourcing challenge solving.
Lecture 36
Build Behavioral Bot Classifier
Assemble a small model pipeline that classifies controlled human and bot traces for later evaluation.
Lecture 37
Behavioral Analysis Report
Summarize model behavior, dominant features, blind spots, and how you would productionize or reject the approach.