Isolation Forest
isolation-forest
linkedin/isolation-forest is a distributed Scala/Spark isolation forest implementation I built at LinkedIn and open sourced for large-scale unsupervised anomaly detection. The newest major update adds Extended Isolation Forest support for random hyperplane splits.
Core Project
GitHub: linkedin/isolation-forest
I built and open sourced this distributed Scala/Spark isolation-forest implementation for large-scale unsupervised anomaly detection at LinkedIn.
Artifacts
Maven Central: com.linkedin:isolation-forest
Published artifacts for integrating the library into JVM-based data pipelines.
Artifacts
PyPI: isolation-forest-onnx
Python package for converting LinkedIn's isolation-forest model format into ONNX for portable inference.
Major Updates
Mar 18, 2026
Extended Isolation Forest for Distributed Spark/Scala Anomaly Detection
March EIF work for linkedin/isolation-forest: random hyperplane splits, validation plots, benchmarks, Spark ML integration, and v4.1.0 support.
Sep 23, 2024
Announcing ONNX Support in Isolation Forest
Details on ONNX export support and deployment options beyond Spark batch inference.
Aug 13, 2019
Open Source: Spark/Scala Isolation Forest Library
Original project announcement and context on anti-abuse production use cases.
External Writing
LinkedIn Engineering
Open Sourcing Isolation Forest
Engineering write-up on motivations, architecture, and applications.
LinkedIn Pulse
Announcing ONNX Support in LinkedIn's Open-Source Isolation Forest Library
Overview of ONNX model export and expanded serving patterns.
Videos
2020
Spark+AI Summit: Preventing Abuse Using Unsupervised Learning
Detection of abusive activity on a large social network is an adversarial challenge with quickly evolving behavior patterns and imperfect ground truth labels. These characteristics limit the use of supervised learning techniques, but they can be overcome using unsupervised methods. To address these challenges, we created a Scala/Spark implementation of the isolation forest unsupervised outlier detection algorithm; we recently open sourced this library (github.com/linkedin/isolation-forest).
2019
Fighting Abuse @Scale: Preventing Abuse Using Unsupervised Learning
Detection of abusive activity on a large social network is an adversarial challenge with quickly evolving behavior patterns and imperfect ground truth labels. These characteristics limit the use of supervised learning techniques, but they can be overcome using unsupervised methods. To address these challenges, we created a Scala/Spark implementation of the isolation forest unsupervised outlier detection algorithm; we recently open sourced this library (github.com/linkedin/isolation-forest).