TRAINING
Forward Pass
LOSS
2.4000
ACC
12.0%
EPOCH
0
lr: 3.0e-4∇: 1.200
RSI TECH
PROGRAM | Security Analytics | 2019 -> 2022

Security Ops: User Behavior Analytics & Anomaly Detection

Large-scale telemetry -> behavioral baselines -> explainable anomaly signals for SOC workflows.

PySparkPythonUnsupervised learningFeature engineeringLog pipelines
  • Processed enterprise telemetry streams (directory, proxy, endpoint) at scale.
  • Unsupervised detection to surface meaningful anomalies with controllable false positives.
  • Analyst-in-the-loop iteration and methodology docs to keep the system auditable.

Context

  • Security teams need behavioral signals they can trust, not black-box alerts.
  • Data volume is high; latency and cost constraints matter.
  • Explainability is mandatory: an alert must be debuggable.

What we built

  • Ingestion and normalization pipelines for multi-source logs.
  • Feature extraction for user/session/device baselines and temporal patterns.
  • Unsupervised scoring and thresholds designed for SOC triage and investigation.

Engineering choices

  • Spark-first design to keep throughput stable under growth.
  • Signals designed to degrade gracefully when data is sparse or partially missing.
  • Documentation treated as a deliverable (how signals behave, why they trigger).
RELATED