Anomaly Detection in Network Logs Using Generalized Isolation Forest and ExIFFI

Explore a comprehensive conference talk that demonstrates a cutting-edge hybrid machine learning approach for detecting anomalies in network logs using Generalized Isolation Forest combined with ExIFFI (Explainable Isolation Forest Feature Importance). Learn how this innovative technique addresses the growing sophistication of cyber threats by efficiently identifying rare and subtle anomalies in high-dimensional network log data while providing transparency through explainable AI. Discover the practical implementation of Isolation Forest algorithms for unsupervised anomaly detection, understand how ExIFFI enhances interpretability by revealing why specific logs are flagged as anomalous, and examine real-world deployment strategies using Apache Spark, Apache Kafka, MongoDB, and Prometheus/Grafana for processing and visualizing network security insights. Gain insights into benchmarking results, hyperparameter tuning methodologies, and lessons learned from handling noisy log data in production environments, with practical guidance on building real-time anomaly detection pipelines that can adapt to modern cybersecurity challenges including Advanced Persistent Threats and polymorphic malware.