Benchmark RadarRelational Anomaly Detection Benchmark
The Relational Anomaly Detection Benchmark evaluates anomaly detection in multi-table relational databases, spanning three settings: LANL cybersecurity events, Amazon user churn, and H&M user churn. It provides standardized tasks and evaluation protocols for relational anomaly detection methods.
- Released
- 2026-08-24
- Readiness
- Runnable
- Primary field
- Cybersecurity
Why it matters
The benchmark addresses the gap in evaluating anomaly detection that preserves relational structure, which is common in real-world databases. It provides a common ground for comparing methods and advancing research in relational learning.
Motivation
Anomaly detection is often applied to data stored in relational databases, yet most existing methods require flattening multiple tables into a single feature matrix.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.