detecting-aws-cloudtrail-anomalies

# Detecting AWS CloudTrail Anomalies ## Overview AWS CloudTrail records API calls across AWS services. This skill covers querying CloudTrail events with boto3's `lookup_events` API, building statistical baselines of normal API activity, detecting anomalies such as unusual event sources, geographic anomalies, high-frequency API calls, and first-time API usage patterns that indicate compromised credentials or insider threats. ## Prerequisites - Python 3.9+ with `boto3` library - AWS credentials with CloudTrail read permissions (cloudtrail:LookupEvents) - Understanding of AWS IAM and common API patterns - CloudTrail enabled in target AWS account (management events at minimum) ## Steps ### Step 1: Query CloudTrail Events Use boto3 CloudTrail client's lookup_events to retrieve recent API activity with pagination. ### Step 2: Build Activity Baseline Aggregate events by user, source IP, event source, and event name to establish normal behavior patterns. ### Step 3: Detect Anomalies Flag unusual patterns: new event sources per user, first-time API calls, geographic IP changes, high error rates, and sensitive API usage (IAM, KMS, S3 policy changes). ### Step 4: Generate Detection Report Produce a JSON report with anomaly scores, top suspicious users, and recommended investigation actions. ## Expected Output JSON report with event statistics, baseline deviations, anomalous users/IPs, sensitive API calls, and error rate analysis. @ref(api-reference.md)