Segmented Billing Anomaly Detection for Cloud Fraud Prevention

The patent segments the computing platform into multiple billable segments (I/O segments, compute segments, storage segments, network segments) and bills customers based on actual usage of each segment. This segmentation approach simplifies the charging mechanism by breaking down complex resource usage into discrete, measurable units while maintaining billing accuracy for each segment independently.

The patent implements anomaly detection mechanisms that continuously monitor billing data against established baselines and thresholds before fraudulent charges are finalized. By performing preliminary anomaly detection on billing anomalies (such as unexpected spikes in resource usage or inconsistent pricing patterns), the system can identify and prevent fraudulent billing attacks before they cause significant financial loss, while maintaining efficient processing through automated threshold-based detection.

The patent applies different monitoring and analysis intensities to different segments based on their fraud risk profiles and business importance. High-value segments with higher fraud risk undergo more comprehensive anomaly detection and simulation checks, while lower-risk segments use simpler monitoring approaches. This local quality differentiation maintains high fraud detection accuracy for critical segments while reducing overall computational overhead by avoiding uniform intensive monitoring across all segments.